Subscribe to Liftr Insights Newsletter

Liftr Insights, a pioneer in market intelligence driven by unique data, has Lambda Labs on its list of cloud service providers.

Lambda, like CoreWeave and Vultr, is a smaller cloud service provider compared to AWS, Azure, and Google Cloud, but is important in the rentable GPU space. With the explosive interest in AI over the past years, smaller cloud providers have taken a prominent role in meeting the high demand for instances such as the NVIDIA Hopper H100 and the B200.

Like CoreWeave, Lambda was among the first providers to offer H100 instances. More recently, the Liftr data showed Lambda as one of the first service providers to add instances based on NVIDIA GH200 Grace Hopper accelerators.

Lambda has closed $320 million in Series C funding to accelerate its presence as a GPU cloud provider, giving the company a $2 billion valuation at closing.

"In addition to seeing the growth in Lambda, as well as the first appearance of NVIDIA's GH200, our Liftr data shows changes in Lambda's GPU pricing," says Tab Schadt, CEO of Liftr Insights. "This is important as companies consider the financial implications for the AI services they are expected to deliver."

As with Vultr, Lambda is included in the Liftr® Intelligence Compute Tracker℠ data service. With over 6 years of trends and comparable history, this data tracks the pulse of AI introductions, directional trends, market share, underlying product details, and insights into the who/what/where of AI infrastructure.

This data set leverages select AI training and AI inference accelerator data from Liftr® Cloud Components Tracker℠. The data layers on AI reservations, previews, and limited access types. In addition to data from AWS, Aliyun Cloud, Azure, Google Cloud, Oracle Cloud, and Tencent Cloud, the data set also includes data from CoreWeave, Vultr, and Lambda.

"There is immense interest in what will happen on the horizon," says Schadt. "Liftr data will provide the insight that companies need for their planning."

For example, the appearance of the recently announced Blackwell B200 GPU instances by NVIDIA is of high interest. In addition to the major cloud service providers named by NVIDIA, Lambda announced that it will be among "the first" service providers to offer NVIDIA's next-generation Blackwell B200 GPUs.

About Liftr Insights

Liftr Insights generates reliable market intelligence using unique data, including details about configurations, components, deployment geo, and pricing for:

Server processors: Intel Xeon, AMD EPYC, Aliyun Yitian, AWS Graviton, and Ampere Computing Altra

Datacenter compute accelerators: GPUs, FPGAs, TPUs, and AI chips from NVIDIA, Xilinx, Intel, AMD, AWS, and Google

As shown on the Liftr Cloud Regions Map at https://bit.ly/LiftrCloudRegionsMap, among the companies tracked are Amazon Web Services, Microsoft Azure, Alibaba Cloud, Google Cloud, Oracle Cloud, Tencent Cloud, CoreWeave, Lambda, and Vultr as well as semiconductor vendors AMD, Ampere, Intel, and NVIDIA. Liftr Insights subject matter experts translate company-specific service provider data into actionable alternative data.

Contact@LiftrInsights.com

Liftr and the Liftr logo are registered service marks of Liftr Insights. The following are trademarks and/or service marks of Liftr Insights: Liftr Insights, Cloud Components Tracker, Intelligence Compute Tracker, and Liftr Cloud Regions Map.

The following are registered intellectual property marks, trademarks, or service marks of their respective companies:

Amazon Web Services

Microsoft Azure

Alibaba Cloud

Google Cloud

Oracle Cloud

Tencent Cloud

CoreWeave

Lambda

Vultr

Intel Corporation

Ampere Computing

NVIDIA

AMD

ARM

SOURCE Liftr Insights

Subscribe to Liftr Insights Newsletter

Subscribe to Liftr Insights Newsletter

Liftr Insights, a pioneer in market intelligence driven by unique market intelligence data, tracks Vultr AI with the rest of the cloud service providers it has data on.

As seen by the expansive growth of NVIDIA, artificial intelligence has become an important factor in market intelligence for all types of companies, including organizations that didn't have AI on their radar at the start of 2023.

While the traditional cloud service providers in the Liftr Insights dataset, such as AWS, Azure, and Google Cloud, still dominate AI discussions, a proper evaluation of the AI infrastructure landscape is not complete without considering some of the smaller providers in the space. Vultr is one such company.

Vultr started in 2014 and is currently the world's largest privately held cloud service provider, but has gained recognition as a provider of the latest versions of NVIDIA GPUs, particularly the NVIDIA Hopper H100 and GH200.

"The cloud space was already interesting," says Tab Schadt, CEO of Liftr Insights, "but the growth of many small players (relative to the size of the major providers we track) is a notable dynamic. Companies like Vultr and CoreWeave are important pieces of the puzzle."

Vultr provides valuable information about clouds and semiconductors: how fast they grow, what is gaining traction (or not), and how quickly they add new types. They have already demonstrated how they can move faster than the larger providers when it came to the deployment of the NVIDIA H100's back in 2023.

"Data on Vultr and CoreWeave help our clients measure the growth of H100 in an objective way," says Schadt. "Market intelligence needs to be watching companies like Vultr for the appearance and growth of other GPUs, such as AMD's MI300 or Intel's Gaudi 3."

Vultr is part of the Liftr® Intelligence Compute Tracker℠ data service. With over 6 years of trends and comparable history, this data tracks the pulse of AI introductions, directional trends, market share, underlying product details, and insights into the who/what/where of AI infrastructure.

This data set leverages select AI training and AI inference accelerator data from Liftr® Cloud Components Tracker℠. The data layers on AI reservations, previews, and limited access types. In addition to data from AWS, Aliyun Cloud, Azure, Google Cloud, Oracle Cloud, and Tencent Cloud, the data set also includes data from CoreWeave data and Vultr.

About Liftr Insights:

Liftr Insights generates reliable market intelligence using unique data, including details about configurations, components, deployment geo, and pricing for:

Server processors: Intel Xeon, AMD EPYC, Aliyun Yitian, AWS Graviton, and Ampere Computing Altra

Datacenter compute accelerators: GPUs, FPGAs, TPUs, and AI chips from NVIDIA, Xilinx, Intel, AMD, AWS, and Google

As shown on the Liftr Cloud Regions Map at https://bit.ly/LiftrCloudRegionsMap, among the companies tracked are Amazon Web Services, Microsoft Azure, Alibaba Cloud, Google Cloud, Oracle Cloud, Tencent Cloud, CoreWeave, and Vultr as well as semiconductor vendors AMD, Ampere, Intel, NVIDIA, and Xilinx. Liftr Insights' subject matter experts translate company-specific service provider data into actionable alternative data.

Liftr and the Liftr logo are registered service marks of Liftr Insights. The following are trademarks and/or service marks of Liftr Insights: Liftr Insights, Cloud Components Tracker, Intelligence Compute Tracker, and Liftr Cloud Regions Map.

The following are registered intellectual property marks, trademarks, or service marks of their respective companies:

Amazon Web Services

Microsoft Azure

Alibaba Cloud

Google Cloud

Oracle Cloud

Tencent Cloud

CoreWeave

Lambda

Vultr

Intel Corporation

Ampere Computing

NVIDIA

AMD

ARM

SOURCE Liftr Insights

Subscribe to Liftr Insights Newsletter

Subscribe to Liftr Insights Newsletter

Since the explosive interest in ChatGPT by OpenAI, the demand for AI chips, especially the NVIDIA Hopper H100, has skyrocketed, but the impact extends to chips by Intel, AMD, and cloud service providers like AWS and Google, themselves.

Liftr Insights' Liftr Intelligence Compute Tracker is hyper-focused on hyperscale AI for market intelligence analysts. This new data set tracks the pulse of AI introductions, directional trends, market share, underlying product details, and insights into the who, what, and where of AI infrastructure.

"Among all the hype and speculation, Liftr data—especially our Intelligence Compute Tracker data—provides objective insights into AI activity," says Tab Schadt, CEO of Liftr Insights. "Like our other offerings, this data is a regular delivery of reliable, objective data."

Liftr Insights cultivated the Intelligence Compute Tracker to focus on data that is driving AI. The data leverages a subset of the Liftr Cloud Components Tracker, which Liftr has offered to corporate and investment clients for the past 6 years. The Intelligence Compute Tracker service expands the data set with supplemental objective data that provides additional insight into infrastructure components that are not fully deployed or have limitations on usage.

CoreWeave and Lamdba, among others, are included in the data set. Both companies received attention for their early deployment of NVIDIA GPUs, especially the H100.

In the coming months, market intelligence analysts will be focused on the arrival of many more chips designed for AI Training and AI Inference. Analysts are watching for the NVIDIA H400, AMD MI300, and the Intel Gaudi 3. The data set also includes accelerators unique to cloud providers such as Google TPUs and AWS Inferentia. This new data set will track when and where these instances appear and where they succeed (or don't).

Historical trends are a critical measure for evaluating current trend rates. The Liftr Intelligence Compute Tracker fully tracks (not just samples) over 75% of the public cloud data with 6 years of history. For example, Liftr customers can see important correlations between the growth of H100's compared to the A100s. That historical trend can help refine projection models.

"It's not just a matter of when we first see accelerators appear," says Schadt. "The more interesting question is what happens after they appear. Do they grow? Flatten out? These all provide insights for designing new products or marketing existing offerings."

About Liftr Insights

Liftr Insights generates reliable market intelligence using unique data, including details about configurations, components, deployment geo, and pricing for:

Server processors: Intel Xeon, AMD EPYC, Aliyun Yitian, AWS Graviton, and Ampere Computing Altra

Datacenter compute accelerators: GPUs, FPGAs, TPUs, and AI chips from NVIDIA, Xilinx, Intel, AMD, AWS, and Google

As shown on the Liftr Cloud Regions Map at https://bit.ly/LiftrCloudRegionsMap, among the companies tracked are Amazon Web Services, Microsoft Azure, Alibaba Cloud, Google Cloud, Oracle Cloud, and Tencent Cloud as well as semiconductor vendors AMD, Ampere, Intel, NVIDIA, and Xilinx. Liftr Insights subject matter experts translate company-specific service provider data into actionable alternative data.

Liftr and the Liftr logo are registered service marks of Liftr Insights. The following are trademarks and/or service marks of Liftr Insights: Liftr Insights, Cloud Components Tracker, Intelligence Compute Tracker, and Liftr Cloud Regions Map.

Visit Liftr Insights

The following are registered intellectual property marks, trademarks, or service marks of their respective companies:

Amazon Web Services

Microsoft Azure

Alibaba Cloud

Google Cloud

Oracle Cloud

Tencent Cloud

CoreWeave

Lambda

Intel Corporation

Ampere Computing

NVIDIA

AMD

ARM

SOURCE Liftr Insights

Subscribe to Liftr Insights Newsletter

Subscribe to Liftr Insights Newsletter

How One Person Can Drive Enormous Change

Liftr Insights has been tracking a wide range of AI and ML projects. Among them, PyTorch Image Models (Timm) stands out. With over 160 million downloads in February 2025, Timm has emerged as the most popular computer vision library. That figure marks a 199% increase over December 2024, highlighting a dramatic surge in adoption. But what’s driving this rapid acceleration?

PyTorch Surges as TensorFlow Declines

Timm outpaced the second most downloaded computer vision library by more than 2-to-1.

Timm is the author of a collection of vision-based deep learning libraries built on the PyTorch ecosystem. PyTorch was released as a suite of vendor-neutral, open-source tools designed to work together. PyTorch, supported by AWS, Microsoft, and even Google, has gained popularity as a more flexible alternative to TensorFlow, which Google released in 2015. Despite being open source, TensorFlow has lost momentum, especially in computer vision use cases.

In recent data, TensorFlow accounted for only a handful of downloads (i.e., 383) compared to just short of 400,000,000 of models with PyTorch included.

Timm has successfully ridden the wave of PyTorch’s growth. In fact, it outpaced the second most downloaded computer vision library(from Meta/Facebook) by more than 2-to-1. Even more remarkable is that Timm is an open-source project created and maintained by a single software engineer, Ross Wightman. Its vendor-neutral nature has enabled Ross and community contributors to prioritize user needs over business goals.

Timm’s Contributions Are Scaling with Usage

Neutrality and community focus aren’t the only reasons for Timm’s success. Sustained momentum also requires frequent updates and a deep alignment with developer needs—an area where Timm shines.

As of February 2025, Timm has authored 1,595 models, a 3.2% increase from the previous month.

The sheer volume and quality of contributions have helped stand apart. On average, Timm models receive over 1,500% more downloads than other computer vision models, and their median downloads are over 3,000% higher.

February 2025 Computer Vision Comparison

Standout Models Powering Growth

Among Timm’s extensive model collection, a few key models dominate usage:

• timm/mobilenetv3_small_100.lamb_in1k

• timm/resnet50.a1_in1k

• timm/resnet18.a1_in1k

These three alone account for over 150 million of February’s 160+ million downloads. Notably, they were created back in 2022 and 2023, before the explosion of mainstream AI interest—making their sustained dominance especially impressive.

Timm Makes Inferencing Scalable and Affordable

While training costs can be significant, Timm supports efficient inference at scale—even on low-power devices. For example:

• ResNet-based models can run on devices like the NVIDIA Jetson Nano, or modern Mac/PCs with acceleration.

• Cloud inference costs (based on Liftr data) on AWS for as low as $0.63 per hour to over $100 per hour, depending on instance type and performance needs, with an global average of $3.95 per hour.

Open-Source Popularity Influences Enterprise Decisions

Popularity attracts a large pool of experienced developers, accelerating time to production.

Projects like PyTorch and Timm are:

• Low-risk and widely adopted: Open-source status ensures community longevity and protection from vendor lock-in.

• Highly integrable: Timm's compatibility with other ML tools increases enterprise agility.

• Community-supported: Its popularity attracts a large pool of experienced AI developers, accelerating time to production.

By contrast, libraries developed within a corporate silo (like some Meta-led projects) may lack the same broad utility and developer trust.

Conclusion

One of open source's greatest strengths is that a single contributor or small group can drive outsized industry impact. While big corporations dominate headlines, stories like Ross Wightman’s Timm project show that individuals still shape the AI landscape in profound ways. Timm is a great example of this phenomenon, and its enormous popularity and adoption are helping developers of all kinds leverage computer vision applications in many industries.

Liftr Insights will continue to keep an eye on AI for intriguing points of note and the data behind them. The next article in this Liftr AI series, which will uncover hidden insights about another critical AI model trend. You can also explore our earlier pieces on AI infrastructure and semiconductor trends, including our series about ARM.

To stay current and informed and to be the first to know about the latest reports like these, subscribe to the Liftr Newsletter.

Subscribe to Liftr Insights Newsletter

Subscribe to Liftr Insights Newsletter

Leveraging their domain experience, Liftr helps data centers forecast their power needs

Liftr Insights, a pioneer with over 6 years of market intelligence experience, offers power usage insights for data center analysts.

For example, NVIDIA H100 chips require more power than an average household, and warnings about the growing demand to support AI computing power requirements. Each of the GPUs driving AI Training and AI Inference requires large amounts of power to produce the high workload output. And hundreds of thousands more GPUs are on their way into data centers.

Liftr® is using its domain expertise for sizing and scoping data centers. Knowing what the public cloud providers (and niche AI providers) are offering is valuable information for consultants attempting to maximize the output for private data centers.

"Our services portfolio is a key part of analyst considerations," says Tab Schadt, CEO of Liftr Insights. "Companies and consultants can leverage the millions of dollars in research that the public cloud providers have invested in, where they spend their AI funds."

Since OpenAI helped inspire the age of AI with the release of ChatGPT, the demand for AI will only continue to grow. With that growth, more chips will be needed to meet those demands. While the performance per chip has and will continue to improve, understanding the power needs is a critical component for any cost and design analysis.

For example, Liftr can show a customer that their footprint for H100's might require 828 kW for 100 servers even before factoring other relevant costs. But, several variables can alter those numbers as Liftr works with consultants on their specific configurations and restrictions.

"By extending our information and domain knowledge into the power and scoping side of the equation, we've opened new doors for our clients," says Schadt. "As with AI, this is only the beginning. More powerful insights still to come!"

Subscribe to Liftr Insights Newsletter

About Liftr Insights

Liftr Insights generates reliable market intelligence using unique data, including details about configurations, components, deployment geo, and pricing for:

Server processors: Intel Xeon, AMD EPYC, Aliyun Yitian, AWS Graviton, and Ampere Computing Altra

Datacenter compute accelerators: GPUs, FPGAs, TPUs, and AI chips from NVIDIA, Xilinx, Intel, AMD, AWS, and Google

As shown on the Liftr Cloud Regions Map at https://bit.ly/LiftrCloudRegionsMap, among the companies tracked are Amazon Web Services, Microsoft Azure, Alibaba Cloud, Google Cloud, Oracle Cloud, Tencent Cloud, CoreWeave, Lambda, and Vultr as well as semiconductor vendors AMD, Ampere, Intel, and NVIDIA. Liftr Insights subject matter experts translate company-specific service provider data into actionable alternative data.

Liftr and the Liftr logo are registered service marks of Liftr Insights. The following are trademarks and/or service marks of Liftr Insights: Liftr Insights, Cloud Components Tracker, Intelligence Compute Tracker, and Liftr Cloud Regions Map.

The following are registered intellectual property marks, trademarks, or service marks of their respective companies:

Amazon Web Services

Microsoft Azure

Alibaba Cloud

Google Cloud

Oracle Cloud

Tencent Cloud

CoreWeave

Lambda

Vultr

Intel Corporation

Ampere Computing

NVIDIA

AMD

ARM

SOURCE Liftr Insights

Subscribe to Liftr Insights Newsletter

🎯 Tired of the Endless Content Struggle?

Let’s face it: constantly creating fresh, engaging content across multiple platforms can be exhausting. Social media demands never slow down, and as creators, entrepreneurs, or marketers, you’re always racing against time. But what if there was an easier, smarter way? Say hello to Blotato, the revolutionary AI-powered content engine designed to turn content overwhelm into content mastery!

Get Blotato

🌟 Meet Your New AI Sidekick: Blotato!

Imagine effortlessly transforming one piece of content—be it a podcast, YouTube video, blog post, newsletter, or even a lengthy PDF—into dozens of unique, tailored posts for LinkedIn, Instagram, TikTok, YouTube Shorts, and more. Blotato makes it possible to generate, remix, and distribute over 50 engaging posts every single week, all customized for each specific platform. No more burnout, no more creative blocks—just continuous, high-quality content at your fingertips.

🎥 Faceless Video Creation, Reinvented!

Camera shy? No worries! With Blotato, you can effortlessly produce engaging, professional-quality videos without ever stepping in front of a camera. Blotato’s AI generates natural-sounding voiceovers, subtitles, and attention-grabbing visuals. Now, anyone can create videos that resonate deeply with audiences, regardless of personal camera comfort or video editing skills.

🎨 Stand Out with AI-Generated Visuals!

Forget about scrolling through endless stock images. Blotato includes powerful AI-driven image generation tools, providing stunning graphics uniquely tailored to your brand and message. Instantly upgrade your visual presence and catch your audience’s eyes every single time.

📅 Never Miss a Beat with Easy Scheduling!

Consistency is key to digital success. Blotato’s intuitive content calendar and built-in scheduling features help you easily plan, preview, and publish your content in advance. This organized workflow ensures your brand stays consistently visible and strategically timed—without the daily stress.

Get Blotato

🤝 Join an Empowering Community of Creators!

Blotato is more than just an AI tool—it’s a supportive creative ecosystem. Founded by content creator Sabrina Ramonov, who grew her following to over half a million and reached 50 million views within six months using Blotato alone, this platform is crafted specifically for creators’ needs. Gain access to weekly office hours, an extensive inspiration database, and a vibrant community sharing tips and motivation. You're never alone in your content journey with Blotato.

⚡️ Why Creators Love Blotato:

• Instantly convert one source into multiple engaging posts.

• Create professional faceless videos with ease.

• AI-generated images that make your brand shine.

• Cross-platform posting simplified.

• Time-saving scheduling to maintain consistency.

• Supportive creator community and regular live Q&A sessions.

Ready to Simplify and Amplify Your Content?

Stop wasting precious hours brainstorming and struggling with content creation. Blotato empowers you to produce high-quality, personalized content effortlessly and consistently. Join thousands of satisfied creators, entrepreneurs, and businesses who have revolutionized their online presence with Blotato.

Embrace the future of content creation—try Blotato today and discover just how easy, effective, and exciting content marketing can be!

Get Blotato

Subscribe to Liftr Insights Newsletter

Liftr Insights data goes back 6+ years with historical trends and highlights increasing prices for accelerators in a hot AI market.

For example, in Q4 of 2023, the average price of accelerators increased for the top cloud providers, includingAWS, Azure, Alibaba (Aliyun) Cloud, and Google Cloud.

When looking at certain NVIDIA GPUs, Azure led the pack with prices increasing by 55.2% in the last few months of Q4 2023. Specifically, these percentages are based on average price for NVIDIA Tensor Core (Ampere) models. Google Cloud was the next largest at 20.0% followed by Alibaba Cloud and AWS at 3.5% and 1.2%, respectively during that same time frame.

A portion of these increases are due to higher core counts, but the prices for these in-demand accelerators also show growth when evaluating the average price per core.

"While we see many valuable data points in the prices across providers and across the globe, Liftr data show that cloud prices don't typically fluctuate to this extent," says Tab Schadt, CEO of Liftr Insights. "It's a sign of the hot AI market. Accelerators—like NVIDIA GPUs—are the technology needed for artificial intelligence training and inference. Demand for AI is driving these price changes."

Among many attributes, Liftr Insights tracks on-demand prices for the top six cloud providers, representing over 75% of the total public cloud space. These prices are measured for all the unique configurations in Liftr Insights data, including numerous dimensions for processors and accelerators, such as global region, architecture, core count, and accelerator count as a few examples.

"This level of information is important for organizations needing to pay attention to their bottom line," says Schadt. "Particularly as they expand into new territories and new technologies, such as artificial intelligence. Our customers will be paying close attention to changes in the data as they occur."

About Liftr Insights

Liftr Insights generates reliable market intelligence using unique data, including details about configurations, components, deployment geo, and pricing for:

• Server processors: Intel Xeon, AMD EPYC, Aliyun Yitian, AWS Graviton, and Ampere Computing Altra

• Datacenter compute accelerators: GPUs, FPGAs, TPUs, and AI chips from NVIDIA, Xilinx, Intel, AMD, AWS, and Google

As shown on the Liftr Cloud Regions Map at https://bit.ly/LiftrCloudRegionsMap, among the companies tracked are Amazon Web Services, Microsoft Azure, Alibaba Cloud, Google Cloud, Oracle Cloud, and Tencent Cloud as well as semiconductor vendors AMD, Ampere, Intel, NVIDIA, and Xilinx. Liftr Insights subject matter experts translate company-specific service provider data into actionable alternative data.

Liftr and the Liftr logo are registered service marks of Liftr Insights. The following are trademarks and/or service marks of Liftr Insights: Liftr Insights, Liftr Cloud Components Tracker, and Liftr Cloud Regions Map.

The following are registered intellectual property marks, trademarks, or service marks of their respective companies:
Amazon Web Services
Microsoft Azure
Alibaba Cloud
Google Cloud
Oracle Cloud
Tencent Cloud
Intel Corporation
Ampere Computing
NVIDIA
AMD
ARM

Subscribe to Liftr Insights Newsletter

Subscribe to Liftr Insights Newsletter

Sort of. But, GPUs aren’t going anywhere

Liftr Insights observed that 37.5% (over 1/3) of AWS’s accelerated servers deployed in the first half of February 2025 were FPGA-based. This deployment supported its recently announced F2 instances. FPGAs are still a small portion (3.3% per Liftr Insights data as of the end of February 2025) of the accelerated computing fleet at AWS. However, the bold investment and launch merit a deeper look at FPGAs in AI and how they can help enterprises in their AI journey.

What is an FPGA?

FPGA stands for field programmable gate array. These chips are deployed in a server and can be repeatedly programmed to perform specific tasks after the manufacturing process. This enables enterprises to define highly specific instructions at the hardware level, providing a high degree of versatility and other benefits, such as high performance for specific workloads. This versatility comes with a price: FPGA-enabled hardware is more expensive and requires specific programming tools and skills. But in the right scenarios, the benefits easily outweigh this cost.

When should enterprises consider FPGA computing in AI?

Ability to customize and test: Because FPGAs are programmable, enterprises can design and deploy specific, higher-performance software workloads. Additionally, the FPGA can be changed over time to meet the solution's needs. This programmability is also very useful in a cloud environment, where a customer can scale up its usage over time and multiple customers can use the same pool of hardware. In the case of AWS F2 instances, Amazon also provides the tooling required to program the FPGAs.

Real-time AI: GPUs are great for batch-based AI operations where a high degree of parallel computing with high data variability is necessary. Training an LLM is a perfect example. However, there are some inference use cases, such as robotics or autonomous vehicles, where GPUs may not be the best choice. In many of those cases, the low latency provided by programming at the hardware level provided by FPGAs is ideal. Other use cases, such as genomics and cybersecurity, have also benefited from FPGAs.

What are the benefits of using FPGA acceleration vs GPUs?

FPGA instances cost $.13 per VCPU per hour vs. $.35 for an A100 GPU in the same US Region.

Lower cost: While FPGAs and GPUs do not necessarily compete, the relative difference in on-demand instance pricing is significant. That’s driven by both instance size and price per VCPU. FPGA instances can be purchased in smaller increments starting at 24 VCPUs versus a high-performance A100 GPU which starts at 96 VCPU on AWS. As tracked in Liftr Insights data, the pricing per VCPU is also lower for FPGA instances. FPGA instances cost $.13 per VCPU per hour vs. $.35 for an A100 GPU in the same US Region. So, in cases well-suited to FPGAs, FPGAs provide a much lower ongoing operating cost.

Lower energy use: The energy requirements of an FPGA-based server tend to be less than a GPU-based system. While there is no hard set of rules regarding the difference, some lab benchmarks have estimated the difference between 10% and 30%¹ in terms of electricity consumption. For those with sustainability requirements, this may be a consideration particularly in FPGA-specific use cases.

Conclusion

In the right circumstances, FPGAs provide a great deal of benefit in delivering some AI use cases. And while GPUs will continue their dominant position for the foreseeable future, enterprises need to consider alternatives depending on the specific needs of the solution. This is where insights and analytics from Liftr Insights can provide a great advantage to analysts using market intelligence to make better data-based decisions for cloud and AI.

Speaking of AI, check out the series of articles about AI models. And, learn more about the growing influence of ARM in the cloud. Stay on top of the latest updates by signing up for the Liftr Newsletter.

Subscribe to Liftr Insights Newsletter

Liftr Insights runs the calculations for 2,000 NVIDIA cards based on its data covering semiconductors and power

AUSTIN, Texas, Aug. 20, 2024 /PRNewswire/ -- Liftr Insights, a pioneer in market intelligence driven by unique data, demonstrated the on-going costs for running popular AI workloads from NVIDIA.

NVIDIA has been dominating news and the markets for its semiconductors. Their latest semiconductors, such as the Hopper H100, are considered essential to the rising demand for AI training and other artificial intelligence processes. Liftr data show that even their earlier types like the A100 remain in high demand.

For example, for a $33M investment in AI accelerator components and $2 M in electric consumption per year, a company could be running 1,000 H100s and 1,000 A100s in Dallas, Texas, which combined could be generating high performance in excess of 44.7 FP64 petaflops.

In comparison, Liftr Insights shows that running the same infrastructure in Houston would generate a larger power bill at approximately $2.1M per year. Annual costs would be less in San Antonio or Austin, at $1.9M and $1.6M, respectively.

"Despite the news of the delay in the Blackwell processes," says Tab Schadt, CEO of Liftr Insights, "major cloud providers like AWS, Azure, and GCP have been increasing their adoption of the latest NVIDIA semiconductors."

Liftr Insights, which tracks the top 6 cloud providers and 3 trending providers, shows the adoption growth of the Hopper brand. The H100 is the most widely discussed model within the Hopper series.

In addition to showing common configurations as well as adoption trends by the major consumers, Liftr data can provide deeper insight for data center operators.

"We help our customers understand the impact of these new chips on output and their financials," says Schadt. "When looking at new AI, it's more than knowing what's available. Rather, for specific configurations, it's understanding the performance, power consumption, and ultimately, the bottom line of on-going costs."

About Liftr Insights
Liftr Insights generates reliable market intelligence using unique data, including details about configurations, components, deployment geo, and pricing for:

• Server processors: Intel Xeon, AMD EPYC, Aliyun Yitian, AWS Graviton, and Ampere Computing Altra

• Datacenter compute accelerators: GPUs, FPGAs, TPUs, and AI chips from NVIDIA, Xilinx, Intel, AMD, AWS, Google, and Qualcomm.

  

As shown on the Liftr Cloud Regions Map at https://bit.ly/LiftrCloudRegionsMap, among the companies tracked are Amazon Web Services, Microsoft Azure, Alibaba Cloud, Google Cloud, Oracle Cloud, Tencent Cloud, CoreWeave, Lambda, and Vultr as well as semiconductor vendors AMD, Ampere, Intel, Qualcomm, and NVIDIA. Liftr Insights subject matter experts translate company-specific service provider data into actionable alternative data.

Liftr and the Liftr logo are registered service marks of Liftr Insights. The following are trademarks and/or service marks of Liftr Insights: Liftr Insights, Cloud Components Tracker, Intelligence Compute Tracker, and Liftr Cloud Regions Map.

The following are registered intellectual property marks, trademarks, or service marks of their respective companies:

Amazon Web Services
Microsoft Azure
Alibaba Cloud
Google Cloud
Oracle Cloud
Tencent Cloud
CoreWeave
Lambda
Vultr
Intel Corporation
Ampere Computing
Qualcomm
NVIDIA
AMD
ARM

SOURCE Liftr Insights

Tabular data models demonstrate an unsung portion of AI model universe by managing complex data

Part 2 of the AI Model Series, by Liftr Insights

There are models downloaded 8x more, on average, than the traditional types of models you know. And, no, it's not a model from DeepSeek, OpenAI, Google, X, or Meta.

Our first article about AI discussed how recent AI events from China challenge the mainstream understanding of AI. Another prevalent misconception, especially outside of the technology space, is that generative AI encompasses the entire AI category of models, and that is not the case. For instance, one of the most popular types of open-source AI models gets eight times as many downloads as the average open-source natural language processing ("NLP") model. Interestingly, most people have never heard of it. It doesn't appear on hot AI trends lists. In fact, there are over 750 NLP models for every one of these types of models.

But if you have massive data sets (millions of entries), this type of model is likely an indispensable tool for your team. Let's dig into Tabular Machine Learning.

What are Tabular Machine Learning Models?

It's not just about what to do with the data, it's about detecting patterns from millions of rows of data.

During the past ten years, enterprises have commonly complained about making sense of the volume of data they have collected. This data could include web analytics, machine logs, and transaction histories. It's not just about what to do with the data, it's about detecting patterns from millions of rows of data. Predictive maintenance is a good example, which helps with the sustainability of machines or other building systems. Humans are unlikely to be able to process data sets of this size easily. This is where Tabular data models add value.

Enterprise use cases for Tabular Models

Tabular models are horizontal in nature and are designed to work well with structured data sets of any kind. This level of versatility is a big benefit since Tabular Models can be employed for many use cases, including:

• Predictive Maintenance

• Customer Segmentation

• Customer value and churn forecasting

• Fraud Detection

• Drug discovery and trail reporting

Small Tabular Models are very powerful

In the case of Tabular Models, Chronos T5 Small and T5 Tiny were the two most popular versions as of December 2024, according to Liftr Insights data.

As we mentioned in last post about DeepSeek, AI models do not have to be big to be powerful. Like large language models, Tabular models come in different sizes depending on the features and performance of the task. In the case of Tabular Models, Chronos T5 Small and T5 Tiny, both products of Amazon, were the two most popular versions as of December 2024, according to Liftr Insights data. These models are great because they can process tens of thousands of rows running on a laptop. This is useful for many enterprise scenarios or as a testing and training environment for larger data sets. Offloading testing and training work to local machines without GPUs is a common way to reduce the impact of more expensive and energy-consuming cloud or server instances.

2 Generations of Chronos are the dominant Tabular model

The older generation T5 models are among the 500 most popular models as measured by downloads.

Liftr Insights data show two distinct groups of Chronos models. The older generation T5 models are among the 500 most popular models as measured by downloads. These models have a very rich feature set ideal for data scientists. The newer Bolt models may not be quite as popular (yet) but are designed with ease of use in mind, including support for non-contextual prompting and high-speed analysis common with business analysts and power users.

Chronos models are sneaky popular

Liftr's reports show that the Natural Language Processing category is the most popular, with more than 300,000 distinct models.

One historical selection metric many enterprises use is the aggregate number of downloads an open-source artifact may have. The theory behind this is that downloads are a proxy for real-world adoption and/or enterprise readiness. However, finding that number is more difficult in the world of AI versus an open-source web server or database. Models are often made available across many products, established cloud platforms, and new platforms like Hugging Face, and some of those platforms do not readily share downloads or usage metrics. It prompts a question: are downloads still an appropriate measure? (It's one of the reasons Liftr Insights data include other types of metrics.)

Tabular models have only 1.0% of the total NLP downloads, but the average number of downloads per model is over 24,000.

We believe models should be compared to other models within in the same category. For example, Liftr's reports show that the Natural Language Processing category is the most popular, with more than 300,000 distinct models. However, the average number of downloads per model is about 3,200. Tabular models have only 1.0% of the total NLP downloads, but the average number of downloads per model is over 24,000. In fact, the Tabular model category has the highest number of downloads per model across any category. This does not suggest Tabular models are better than NLP. But, if you are looking at Tabular models, the threshold for what's popular or what a "good" number of downloads should be, it's going to be higher than 24,000.

Liftr Insights: Going beyond the cloud and into AI

At Liftr Insights, we aim to help enterprises and investors make data-driven decisions based on reliable and well-curated market analytics. Liftr has more than half a decade of experience and curated data from the top nine cloud providers. These insights have helped enterprises and investors decide which capabilities fit them, including regional availability, price/performance, and the availability of the right instances.

When considering your AI strategy, it's better to understand the data and the trends than just trusting what the mass market and well-monied vendors tell you. In this blog series, we will discuss some of what we learned with our new AI data set. In our next post, we will discuss how single-contributor AI projects like timm can have an outsized impact in a market dominated by giants.

To stay on top of news about AI like the above (including the next in the series of articles about AI), subscribe to the newsletter today.

Analysts are blinded by DeepSeek and OpenAI and missing the depth of data revealed by Liftr Insights

Part 1 of the AI Model Series, by Liftr Insights

In January 2025, the announcement of a new AI model called DeepSeek R1 significantly impacted U.S. financial markets. This China-developed large language model raised investor alarms due to its performance and the claims that it was built for about $6M USD, compared to the hundreds of millions invested in the more well-known U.S frontier models such as OpenAI and Gemini.

It’s important to state that China’s AI aspirations go beyond DeepSeek. For instance, a Liftr Insights report shows that as of the end of the previous quarter, Alibaba Cloud and Tencent Cloud accounted for 31.4% of all the AI-focused instance types across all the major cloud providers. In contrast, AWS was only 20.2%, which was lower than Alibaba Cloud alone.

The Emergence of DeepSeek

Of the 1.2 million AI and ML models hosted on the Hugging Face hub in December 2024, only 28.4% are tagged as natural language processing.

The emergence of DeepSeek and Chinese AI challenge common AI beliefs. In contrast to some of these misperceptions, Liftr data reports show:

1. It’s not all about Natural Language Processing (NLP). During the past 2 years, NLP and broad-based frontier models have gotten the lion’s share of media, analyst, and investor attention when it comes to AI. But, DeepSeek’s rapid ascent from a field of over a million AI and ML models points to a much bigger and vibrant AI ecosystem. Of the 1.2 million AI and ML models hosted on the Hugging Face hub in December 2024, only 28.4% are tagged as natural language processing. This is obscured by the fact that 61.2% of the downloads were focused on NLP due to some popular models dominating the attention. That means there are about 900,000 other AI and machine learning models helping accomplish different objectives. For example, 5.6% of these other models are focused on computer vision, which we might delve into in a future article.

Among models specifically stated with parameters, Liftr data show that 90.3% of models are trained on (or based on models trained on) a medium size of parameters.

2. You don’t need to be big and spend big to use AI. Having extensive resources may have been true in the past when models were less mature and accurate. Brute force efforts to scale training and inference networks may have been useful for frontier LLMs, but DeepSeek has been trained using a distillation model. This approach “shrinks” models to be very small, efficient, and highly focused on specific tasks.
   
   Like we are seeing with DeepSeek, we also see great results from smaller models from other authors. For example, IBM’s Granite models are available in multiple sizes from 3 billion training parameters up to over 34 billion, depending on the project requirements. Among models specifically stated with parameters, Liftr data show that 90.3% of models are trained on (or based on models trained on) a medium size of parameters. Some small models can even be run on CPU alone or on a GPU-enabled laptop such as a Macbook Pro. Of course, your performance may vary.

3. 

Liftr’s data shows AI specific instances already comprise 6.0% of all workloads on leading clouds.

3. You don't need a cloud to get started. Cloud service providers have been at the forefront of investing in and marketing AI-based solutions. They have a huge incentive to do so. According to Liftr’s cloud data, AI instances can cost 20% to 30% more per hour than CPU-only instances of similar size. This marketing push is working as Liftr’s data shows AI specific instances already comprise 6.0% of all workloads on leading clouds. So, enterprises may have come to believe that the only way to access these models is through a cloud-based deployment. This is not true. DeepSeek, IBM, NVIDIA, and Meta all deploy their models to be deployed in clouds or on-premise in enterprise data centers. And with smaller models available, private deployment is more feasible than ever.

Making sense of a million models

While Hugging Face and other platforms provide a great service for connecting enterprises to models, open-source AI and ML present many of the same challenges as other open-source software projects. Enterprises want some degree of assurance that a model is viable, stable, and has a vibrant set of contributors. Additionally, enterprises should understand whether other models or tools complement or compete with a model.

When considering your AI strategy, it’s better to understand the data and the trends than just trusting what the mass market and well-monied vendors tell you.

Digging Deeper into AI

At Liftr Insights, we aim to help enterprises and investors make data-driven decisions based on reliable and well-curated market analytics. Liftr has six years of processes and experience curating data about the top nine cloud providers, semiconductors, and AI. These insights have helped enterprises and investors decide which capabilities fit them, including regional availability, price/performance, and the availability of the right instances.

When considering your AI strategy, it’s better to understand the data and the trends than just trusting what the mass market and well-monied vendors tell you. In this AI series, we will discuss valuable insights we are seeing in the AI data. In our next article will discuss one of the most prolific machine-learning models and, given the emphasis on Natural Language Programming, you have probably never heard of this type of AI model.

To stay on top of news about AI like the above (including the next in the series of articles about AI), subscribe to the newsletter today.

Contact Liftr Insights

2025 marks a pivotal year for AI video generation, with multimodal generation techniques centered around Diffusion Models achieving groundbreaking advancements. Video Large Language Models (Video LLMs), exemplified by Sora and RunwayML, have successfully tackled industry challenges related to video temporal coherence and physical law simulation through spatiotemporal joint modeling.

This year, the open-source models SkyReels (based on dynamic attention mechanisms) and WAN (Wavelet-Augmented Networks) have significantly improved pixel-level precision and inference efficiency in video generation through modular architecture design and multimodal fusion capabilities. And this is just the beginning of 2025.

Among them, the two open-source models that intrigue people the most are WAN and SkyReels. WAN, backed by Alibaba Group, is a powerhouse product with abundant resources and funding, so its impressive results come as no surprise. However, what’s truly remarkable is that SkyReels—a sleek and elegant platform—open-sourced even before WAN. It’s like DeepSeek(Video Model Version) and has held a spot in the Hugging Face Top 10 for two consecutive weeks.

Open-Source Repositories

• SkyReels-V1

  https://github.com/SkyworkAI/SkyReels-V1

• SkyReels-A1

  https://github.com/SkyworkAI/SkyReels-A1

Key features include:

Video generation

• Cinematic Expression Recognition: 11 types of facial expression understanding for characters in film and drama, including expressions such as disdain, impatience, helplessness, and disgust, with emotional intensity levels categorized into strong, medium, and weak.

• Character Spatial Awareness: Leveraging 3D reconstruction technology to comprehend spatial relationships among multiple characters, enabling cinematic positioning.

• Behavioral Intent Understanding: Constructing over 400 behavioral semantic units for precise action interpretation.

• Scene-Performance Correlation: Analyzing the interplay between characters, wardrobe, setting, and plot.

• Drama Generation: SkyReels Drama Generation is a more professional film and television production tool, which easily enables your text story into a visual master-level film and television storyboard. It supports the adjustment of character costumes and sounds as well.

Visit SkyReels

Empowering Content Creators Worldwide

SkyReels was designed with the modern creator in mind. Whether you’re a social media influencer, a marketing professional, or an independent filmmaker, the platform’s intuitive design and robust feature set streamline the editing process. By reducing the time from concept to publication, creators can now produce and share content more frequently, keeping up with the fast-paced digital landscape.

The benefits extend beyond just speed:

• Enhanced Storytelling: The platform’s smart curation ensures that your most compelling moments are highlighted, creating a more impactful narrative.

• Cross-Platform Readiness: With formats and aspect ratios optimized for platforms like Instagram, TikTok, and YouTube, SkyReels ensures your content looks its best no matter where it’s viewed.

• Community and Learning: SkyReels also nurtures a vibrant community of creators. From engaging social media channels like YouTube (@SkyReels_AI) and X (formerly Twitter) and discord, users can find inspiration, share tips, and witness live demonstrations of how SkyReels is pushing the boundaries of video creation.

Start to create your first high-quality story now! Go to SkyReels.

Press and Industry Recognition

SkyReels has already begun to capture the attention of industry leaders and media outlets. A recent press release featured on Yahoo Finance heralds SkyReels as an “AI-powered reel creator that is set to transform digital content production.” This accolade underlines the platform’s potential to disrupt traditional video editing paradigms and highlights its role as a trailblazer in AI-driven creative technology.

Looking to the Future

As content consumption continues to evolve, so too will the tools that power creative expression. SkyReels is poised to remain at the forefront of this evolution, consistently innovating and expanding its features to meet the changing needs of its users. With regular updates, community feedback, and a commitment to excellence, SkyReels is more than just a product—it’s a partner in every creator’s journey toward captivating storytelling.

HackerRank, a company primarily known for its tech hiring solutions, recently introduced a benchmark designed to challenge AI models and test how well LLMs solve complex software tasks. The initial press release explained, “The ASTRA Benchmark consists of multi-file, project-based problems designed to mimic real-world coding tasks. The intent of the HackerRank ASTRA Benchmark is to determine the correctness and consistency of an AI model’s coding ability in relation to practical applications.”

Seeming to acknowledge the deviation from the company’s other offerings, which focus on screening, interviewing, engaging and skilling developer talent, HackerRank co-founder and CEO Vivek Ravisankar said, “As software development becomes more human + AI, it’s important that we have a very good understanding of the combined abilities. Our experience pioneering the market in assessing software development skills makes us uniquely qualified to assess the abilities of AI models acting as agents for software developers.”

Whether true or not, HackerRank ASTRA enters the market at an interesting moment, with a battle royale raging between AI2, Anthropic, DeepSeek, Google, OpenAI, xAI and others. While the initial release of HackerRank ASTRA does not address all aspects of these models, it does acknowledge the limitations of other benchmarks and seeks to offer a more comprehensive framework, according to the paper published by the HackerRank team.

As the paper explains, “HackerRank-ASTRA is a benchmark built from HackerRank’s proprietary library of multi-file, project-based software development problems. These problems were originally designed to assess the software development skills of human developers across a wide range of skill domains in realistic, project-like settings. We observed that even advanced large language models (LLMs) face significant challenges when solving these problems, which motivated the creation of this benchmark.”

As such, the current version of the open-source benchmark includes:

• Diverse skill domains: Sixty-five product-based coding questions, categorized into 10 primary coding skill domains and 34 subcategories.

• Multi-file project questions: An average of 12 source code and configuration files per question as model inputs. This results in an average of 61 lines of solution code per question.

• Model correctness and consistency evaluation: HackerRank ASTRA prioritizes metrics such as average scores, average pass@1 and median standard deviation.

• Wide test case coverage: HackerRank ASTRA’s dataset contains an average of 6.7 test cases per question to rigorously evaluate the correctness of implementation.

• Leaderboard: Full report and analysis results are available on the HackerRank website.

There are some limitations to the initial benchmark, which HackerRank acknowledged in its research paper. Notably, the benchmark focuses on front-end projects, which means it under-represents back-end skills and other domains. Still, HackerRank is already on top of that issue, with plans for future iterations that will include a broader range of technologies. On top of that, newer AI models are still being added to the leaderboard (about as quickly as they enter the market). Usefulness and application over time are ultimately up to the AI community to determine, but for now, HackerRank points to what the benchmark actually is proving, and that’s how today’s frontier models perform on real software development challenges.

Menu-Order-AI was created to solve a very real, very current problem: people love going out to eat, but they’re ordering differently than they used to. With nearly 9% of the U.S. population now using GLP-1 medications and millions more eating with intention, the traditional restaurant menu no longer supports how diners actually make decisions.

Diners aren’t asking restaurants to change their menus. They still want to enjoy the experience, the atmosphere, and the social moment. What they need is confidence — confidence that what they’re ordering aligns with how they eat right now. That moment of uncertainty happens at the table, not before they leave the house and not after the meal is over.

Unlike calorie trackers or meal planners that work before or after eating, Menu-Order-AI operates in real time. From one restaurant choice or scan of a menu, the app instantly analyzes the menu and highlights options that align with high-protein, GLP-1 friendly, low-carb, low WW point, and plant-based preferences. No guesswork. No jumping between apps. No awkward ordering conversations.

This matters because GLP-1 users aren’t eating less just to eat less — they’re prioritizing quality, protein, and how food makes them feel. Restaurants are already seeing this shift reflected in smaller portions ordered, fewer shared appetizers, and lower alcohol consumption. Industry data shows restaurants losing an estimated 7% of revenue due to these changes, not because people stopped dining out, but because they’re unsure what to order.

By guiding diners toward meals that work for their lifestyle, the app restores confidence at the table. When diners feel confident, they enjoy the experience more — and they’re more likely to return. The solution doesn’t disrupt restaurants, require menu changes, or add friction. It simply supports the diner at the exact moment of decision.

This is why Menu-Order-AI is best described as a real-time dining companion, not a diet app. It’s designed for real life: business dinners, date nights, travel, group meals, and everyday dining — whether eating in or ordering delivery.

As GLP-1 adoption continues to rise and intentional eating becomes the norm, the question is no longer if dining behavior has changed, but how restaurants and diners adapt together. Menu-Order-AI bridges that gap by empowering diners to order with confidence — every time, anywhere.

The app is available now on the Apple App Store and Google Play Store.

Menu-Order-AI

Like most industries, AI and tech have had a profound effect on the legal industry. For years, most legal tech tools focused on making existing processes faster and more organized. However, what’s emerging are more advanced tools that point to a broader shift in how attorneys do their work. Instead of focusing purely on efficiency, legal tech is enabling a process level change where attorneys can rely on technology to support entire legal workflows.

Here are three trends we’re seeing in the legal tech arena as we go into 2026.

Darrow AI

#1 Litigation analytics becomes a core capability

Litigation analytics will function as foundational infrastructure rather than a specialized feature. Earlier generations of analytics concentrated on historical outcomes, but in 2026, analytics will operate continuously and earlier in the lifecycle, shaping how matters are identified and evaluated.

Advances in large scale analysis of litigation activity and public data will enable earlier detection of potential claims and exposures. As new information enters these systems, analytical models will update assessments of case strength and litigation risk throughout the life of a matter. These outputs are likely to influence portfolio construction and capital allocation rather than serve as static reference material.

The same analytical foundation is expected to support adjacent domains such as insurance underwriting and litigation finance. Filed cases will increasingly represent one expression of a broader intelligence layer that informs valuation and risk.

Darrow AI

#2 The industry moves toward agent-based models

Legal AI in 2026 will rely more heavily on agent-based systems capable of executing multistage work with limited human input. Instead of isolated prompt interactions, platforms will deploy coordinated agents that carry out sequences of analysis as conditions evolve.

These agents are likely to operate in parallel across distinct responsibilities. One may handle docket ingestion and normalization. Another may focus on jurisdictional dynamics or precedent analysis. Additional agents may evaluate exposure and downstream implications. Working together, they will support complex legal workflows and enable higher quality decisions at scale.

Agent-based systems align closely with the iterative and context dependent nature of legal work. They’re going to allow human attorneys to focus more on strategy and judgment rather than manual synthesis.

Darrow AI

#3 Embedded intelligence replaces standalone legal tech tools

Legal intelligence is expected to surface contextually within existing systems rather than through standalone dashboards. Insight will appear at the point of decision, embedded directly into operational workflows.

Legal and regulatory risk assessment is likely to become part of routine operations, and legal violations, such as data breaches, will be detectable using public data before products or features reach market.

As intelligence is designed to function across multiple environments, operational friction will decrease while access expands. Firms will rely on fewer platforms and expect deeper analytical value from each. Intelligence will operate as an execution layer integrated into daily work, supporting action rather than requiring deliberate consultation.

Darrow is the leading legal intelligence company identifying legal risk at scale. To learn more about how Darrow uncovers legal violations, check out this article on Darrow’s blog: Using Legal Intelligence to Turn Data Into Justice.For more information about Darrow, visit www.darrow.ai.

⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐
Would you like to be featured in our newsletter🔥 and get noticed, QUICKLY 🚀? Simply reply to this email or send an email to editor@aibuzz.news, and we can take it from there.
⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐

HackerRank, a company primarily known for its tech hiring solutions, recently introduced a benchmark designed to challenge AI models and test how well LLMs solve complex software tasks. The initial press release explained, “The ASTRA Benchmark consists of multi-file, project-based problems designed to mimic real-world coding tasks. The intent of the HackerRank ASTRA Benchmark is to determine the correctness and consistency of an AI model’s coding ability in relation to practical applications.”

Seeming to acknowledge the deviation from the company’s other offerings, which focus on screening, interviewing, engaging and skilling developer talent, HackerRank co-founder and CEO Vivek Ravisankar said, “As software development becomes more human + AI, it’s important that we have a very good understanding of the combined abilities. Our experience pioneering the market in assessing software development skills makes us uniquely qualified to assess the abilities of AI models acting as agents for software developers.”

Whether true or not, HackerRank ASTRA enters the market at an interesting moment, with a battle royale raging between AI2, Anthropic, DeepSeek, Google, OpenAI, xAI and others. While the initial release of HackerRank ASTRA does not address all aspects of these models, it does acknowledge the limitations of other benchmarks and seeks to offer a more comprehensive framework, according to the paper published by the HackerRank team.

As the paper explains, “HackerRank-ASTRA is a benchmark built from HackerRank’s proprietary library of multi-file, project-based software development problems. These problems were originally designed to assess the software development skills of human developers across a wide range of skill domains in realistic, project-like settings. We observed that even advanced large language models (LLMs) face significant challenges when solving these problems, which motivated the creation of this benchmark.”

As such, the current version of the open-source benchmark includes:

• Diverse skill domains: Sixty-five product-based coding questions, categorized into 10 primary coding skill domains and 34 subcategories.

• Multi-file project questions: An average of 12 source code and configuration files per question as model inputs. This results in an average of 61 lines of solution code per question.

• Model correctness and consistency evaluation: HackerRank ASTRA prioritizes metrics such as average scores, average pass@1 and median standard deviation.

• Wide test case coverage: HackerRank ASTRA’s dataset contains an average of 6.7 test cases per question to rigorously evaluate the correctness of implementation.

• Leaderboard: Full report and analysis results are available on the HackerRank website.

There are some limitations to the initial benchmark, which HackerRank acknowledged in its research paper. Notably, the benchmark focuses on front-end projects, which means it under-represents back-end skills and other domains. Still, HackerRank is already on top of that issue, with plans for future iterations that will include a broader range of technologies. On top of that, newer AI models are still being added to the leaderboard (about as quickly as they enter the market). Usefulness and application over time are ultimately up to the AI community to determine, but for now, HackerRank points to what the benchmark actually is proving, and that’s how today’s frontier models perform on real software development challenges.

Over the summer, Darrow, the legal intelligence company built to surface hidden legal risk and turn it into action, detected an anomaly in the code of a fertility tracking app.

On the surface, nothing seemed unusual. The app’s latest update notes mentioned only “general performance improvements” and “minor bug fixes.” But Darrow’s legal intelligence infrastructure flagged something unexpected: the app started transmitting unique device identifiers to an advertising partner that was not listed anywhere in its privacy policy.

If true, this would mean that users’ extremely sensitive reproductive data is unknowingly being tracked and profiled to a third party without consent.

Darrow AI

At first, the signal did not appear concerning. It was the kind of background telemetry many apps use for crash analytics or ad attribution. Still, Darrow’s system marked it as suspicious. The Legal Intelligence team compared the app’s new code signature to its earlier version, then ran the update through a series of checks: changes in requested permissions, modifications to background processes, and differences in outbound network traffic. This revealed a significant change and potential privacy violation: the app had quietly integrated a new module from an advertising software development kit (SDK) that, according to publicly available documentation, could link identifiers with precise geolocation data and trace users’ routines from morning check-ins to evening updates.

Darrow AI

To the average user, this would be invisible. From a legal perspective, however, it raised serious concerns. Analysts reconstructed the data flow, cross-referencing the app’s behavior with recent rulings involving mobile tracking technologies. Combining persistent identifiers and location data and transmitting them to a third party without explicit consent can violate several privacy statutes, including the California Invasion of Privacy Act (CIPA), a state statute that prohibits the unauthorized interception or sharing of communications and data, and often serves as the basis for class action lawsuits.

Darrow automatically flagged the incident as a potential privacy violation and sent it to the company’s privacy intelligence team for legal review. The privacy team began investigating and tracked the timeline of the code changes and confirmed that the new permissions requested access to background location data even when the app was not running. Legal analysts reviewed the information alongside recent court rulings involving unauthorized tracking in mobile applications. The team’s review showed a strong resemblance to earlier cases that resulted in major settlements, particularly those where users had no way to opt out.

Over the following days, Darrow’s legal intelligence analysts mapped the full scope of the issue and analyzed public download data and app store activity to estimate that millions of users had installed the affected version since the update. Each of them likely had their movements and device IDs logged and shared with an undisclosed third party.

Because the data was highly sensitive and the exposure widespread, the incident constituted a serious privacy breach. Darrow’s legal experts identified this as an alleged CIPA violation and translated the findings into a full assessment, including projected damages and settlement scenarios. They assembled a detailed case memorandum with the evidence, legal theories, and impact analysis, and brought it to a leading plaintiff firm. The firm determined the case was viable and filed a putative class action to seek relief for the exposed users.

Through the combination of machine learning, data analysis, and legal expertise, Darrow’s legal intelligence analysts turned a few lines of suspicious code into a potential nationwide enforcement action that could secure relief for millions of people whose most intimate health choices were tracked and sold without their consent.

If successful, the case will send a clear message that exploiting reproductive data for ad targeting carries real financial and legal consequences. This is the power of legal intelligence: turning invisible technical changes into accountability, and shifting power back to the people whose lives and data are on the line.

To learn more about how Darrow uncovers legal violations, check out this article on Darrow’s blog: Using Legal Intelligence to Turn Data Into Justice. For more information about Darrow, visit www.darrow.ai.

Darrow AI

Legal tech is evolving from back office support to frontline enforcement.

While most AI headlines focus on generative tools for coding, writing, or design, a quiet revolution is happening in a very different part of the tech world. Darrow, a legal tech company, has built technology that helps lawyers find legal violations that no one else sees.

This new category of legal tech is called legal intelligence, and it’s the structured process of scanning public data to detect legal violations at scale. Instead of waiting for harm to be reported, carefully built legal intelligence infrastructure finds patterns of misconduct hiding in publicly available sources and builds evidence to turn them into actionable cases.

This is exactly what Darrow is doing. Combining large language models, machine learning, web intelligence, and legal expertise, this tech company is powering a new kind of legal discovery.

Most Data Gets Ignored. Darrow Turns It Into Evidence.

Public data is published every second across the internet: user complaints, app store reviews, environmental filings, financial disclosures, and regulatory records. These documents often sit untouched, even when they contain signs of unlawful behavior.

Darrow’s technology scans data from thousands of sources, across multiple domains, and AI models summarize and organize unstructured text, then cluster similar issues together. The team’s attorneys and legal intelligence analysts then evaluate whether the content might point to a violation of law.

This process blends automation with human judgment. It doesn’t replace lawyers, but it gives them a clearer view of where harm is happening and how to respond.

For example, Chicago-based class action attorney, Kristina Carroll, partnered with Darrow after the company flagged privacy concerns involving the dating apps Bumble and Badoo. Carroll filed a class action against parent company Bumble Inc., which settled last year for $40 million.

What’s fascinating is the fact that the case didn’t come from a whistleblower or a high-profile news investigation, but from public data interpreted by attorneys and a legal intelligence system trained to look for the kinds of issues most people miss.

A Targeted System for Different Types of Harm

Not every legal violation looks the same. A case about data privacy involves different signals than one about environmental harm or deceptive pricing. Darrow’s team of data scientists, engineers, attorneys, and legal intelligence analysts build specialized detection tools for each area of law, combining subject-matter expertise with training data specific to the problem.

The company builds tools with different intent, based on the type of violation it’s looking for. In an antitrust investigation, the company might focus on price changes and customer complaints. In a privacy case, it might flag suspicious onboarding language or metadata collection.

Legal Intelligence Makes the Invisible Visible

Evidence for legal violations is often scattered across unrelated sources, buried in technical documents, or locked behind vague legalese.

Legal intelligence brings that evidence to the surface. It helps lawyers spot systemic problems, evaluate them faster, and take action before harm spreads further.

Darrow is building the infrastructure needed to uncover violations at scale and connect them to real-world enforcement, creating a whole new approach to scaling legal accountability.

To learn more about Darrow’s technology, check out this article on Darrow’s blog: Using Legal Intelligence to Turn Data Into Justice

As the demand for high-speed connectivity continues to grow, mobile operators are struggling to keep up. Their current approach—expanding Radio Access Networks (RANs)—is not sustainable. RANs account for 80% of all mobile network energy consumption, with energy costs outpacing sales growth by over 50%.

Recognizing this growing challenge, Net AI was founded to develop AI-driven solutions that address network efficiency and sustainability. The mission is to bridge the gap between telecommunications and machine learning, enabling operators to optimize the performance of their networks while reducing energy consumption.

Industry Status Quo

Most actors in the telecommunications industry believe that AI systems are not accurate enough in their predictions of traffic demands and would consume too much energy themselves. This stems from a disconnect between the telecom and machine learning (ML) communities.

Traditionally, telecom networks have relied on rigid, rule-based optimization techniques designed for deterministic systems, while ML involves building statistical models based on extensive data used for training. Because of this fundamental difference, many telecom professionals are skeptical about allowing AI to make decisions that could impact network reliability. On the other hand, ML experts often underestimate the strict performance requirements and real-time constraints specific to telecom networks. This disconnect has prevented the industry from fully realizing AI’s potential in optimizing network operations. Bridging this gap requires specialized AI models tailored to telecom environments, ensuring both accuracy and efficiency without compromising service quality.

Meet Net AI

Net AI has been developing domain-specific ML/AI models that can help telecom operators reduce energy consumption and improve network efficiency. Our research has shown that significant energy savings—up to 60%—can be achieved by selectively deactivating parts of the network infrastructure during low-demand periods. However, this must be done carefully to ensure that service quality is not compromised.

Its approach focuses on:

📡 Accurate Traffic Forecasting: Providing real-time insights into network performance and resource usage to enable informed decision-making.

⚡ Efficiency Improvements: Helping operators optimize network resources to manage energy costs while maintaining service quality.

🔎 Proactive Network Management: Identifying unusual patterns in traffic to predict and prevent network incidents.

Our team’s unique expertise at the intersection of AI and telecoms has enabled us to build perhaps the most accurateAI-driven forecasting engine on the market (patent pending). What makes it different?

🚀 Scalability: Our AI is 100 times more scalable than competitors and inherently low in power consumption. We have achieved this via domain-specific neural architecture design.

🎯 Performance: We achieve over 95% accuracy in traffic forecasting. Our data handling pipeline employs pre-processing that improves the sensitivity and generalization ability of trained models.

🔒 Reliability: Our proprietary training methodology ensures that AI models avoid underestimation errors, crucial for maintaining service quality.

By challenging the widely held perception that service disruption risks and computational cost outweigh the benefits of AI-driven forecasting, we’ve created a solution that gives operators confidence to automatically and intelligently switch off targeted network infrastructure while maintaining service quality.

With additional integration work, similar strategies could be applied in data centers, addressing energy efficiency challenges on a broader scale.

Shaping the Future of Telecoms 🌍📶

The telecommunications industry is at a critical juncture where intelligent automation and energy efficiency are becoming key priorities. As AI continues to evolve, its role in infrastructure management will expand, providing new opportunities to enhance service quality while reducing operational costs.

Net AI is playing an active role in this transformation. By combining expertise in AI and telecom, they aim to contribute to a more efficient, intelligent, and sustainable future for network operations. Their work within the ETHER project further supports this goal as it focuses on improving network coverage and energy efficiency through integrated terrestrial and non-terrestrial networks.

Net AI has already demonstrated that AI-driven solutions can play a meaningful role in optimizing telecom infrastructure. As the industry moves forward, we continue to explore new ways to refine and expand these capabilities.

Hey there, AI enthusiast! 👋

Welcome back to AI Buzz! — The AI money machine hit a new gear this week. SpaceX pulled off the largest IPO in human history. Jeff Bezos finally revealed the secretive $41 billion AI startup he’s been building. China’s humanoid robot makers are racing to go public. And as all this capital floods in, the Magnificent Seven just shed $2 trillion in a single month. Buckle up — here’s what you need to know.

Who’s in Charge Here? Human–Agent Teams, Agentic Trading & Molecules as Code

As AI agents start trading, building, and running teams, one question matters most: who’s actually in control? Distill AI breaks it down in their latest issue — a must-read for anyone serious about where AI is headed.

🔥 TODAY’S TOP STORIES

SpaceX Pulls Off the Largest IPO in History — Closes Up 25% at $1.77 Trillion

Why it matters: SpaceX (ticker: SPCX) debuted on the Nasdaq on June 12, raising $75 billion — more than double the previous record (Saudi Aramco’s $35.4B in 2019). Shares priced at $135, opened at $150, and closed at $168.70, a ~25% first-day gain that valued the company at roughly $1.77 trillion. The newly public entity includes xAI, Grok, the X social network, and the Colossus data centers, all folded in via February’s xAI acquisition. Analysts called the orderly debut a green light for OpenAI and Anthropic to proceed with their own fall listings.

💡 The takeaway: A textbook ~25% pop — not a flop, not a frenzy — is exactly the signal the AI IPO pipeline needed. The floodgates for mega-listings are officially open.

Jeff Bezos Unveils Prometheus — a $41B Bet on an “Artificial General Engineer”

Why it matters: Bezos stepped back into a CEO role to reveal Prometheus, his secretive “physical AI” startup, which just raised $12 billion at a $41 billion valuation (backers include Bezos himself, JPMorgan, Goldman Sachs, and BlackRock). Instead of chatbots, Prometheus is building AI that designs and manufactures real-world objects — jet engines, chips, bridges, even drug compounds — aiming to make the “dream-to-build loop” 10x faster. The 150-person team spans San Francisco, London, and Zurich.

💡 The takeaway: While everyone chases artificial general intelligence in software, Bezos is betting the next frontier is the physical world. If it works, it could reshape manufacturing itself.

China’s Humanoid Robot Gold Rush: EngineAI Files for Hong Kong IPO

Why it matters: EngineAI — a three-year-old Shenzhen startup valued at $1.5 billion — filed confidentially for a Hong Kong IPO, just two weeks after opening a factory that builds one humanoid robot every 15 minutes. It’s part of a stampede: sector leader Unitree has filed for a $7 billion IPO, and hand-maker Linkerbot is chasing a $6 billion raise. Roughly $22.6 billion has already been raised in Hong Kong AI and robotics listings this cycle.

💡 The takeaway: Public markets are now willing to fund machines that walk, grip, and patrol. China is moving fast to dominate embodied AI — and the West should be paying attention.

Nvidia’s Vera CPU: A “Side Door” Back Into China

Why it matters: With its top AI GPUs frozen out of China by US export controls (Jensen Huang says market share fell to roughly zero), Nvidia is pitching its new Arm-based Vera CPU to Chinese clients, with orders open now and deliveries as soon as August. The trick: US controls mainly target high-end GPUs, not general-purpose CPUs. One major Chinese cloud provider is already prepping an order for 300+ dual-Vera servers. Nvidia targets $20 billion in Vera revenue by its fiscal year-end.

💡 The takeaway: Export controls created a wall — so Nvidia found a window. Watch whether regulators reclassify Vera once it gains traction in Chinese data centers.

The IPO Hangover: Magnificent Seven Sheds $2 Trillion in a Month

Why it matters: As SpaceX, OpenAI, and Anthropic line up to go public, investors are bracing for a “share avalanche.” The Magnificent Seven (Microsoft, Amazon, Apple, Alphabet, Nvidia, Tesla, Meta) erased roughly $2 trillion in market value this month — more than two-thirds of the S&P 500’s total June decline. The fear: a flood of new mega-IPOs will pull investor dollars out of today’s hot AI trades. Goldman Sachs counters that $1 trillion in buybacks should offset the new supply.

💡 The takeaway: The AI boom giveth and taketh away. As private giants go public, expect capital to rotate violently — and volatility to stay elevated through the IPO wave.

Anthropic vs OpenAI: An Accounting Showdown Could Reshape Their IPOs

Why it matters: As both labs prep confidential IPO filings, a quiet but consequential dispute is surfacing: they count revenue differently. Anthropic recognizes gross revenue (acting as principal with cloud partners), while OpenAI reports net revenue after paying Microsoft. Bank of America estimates Anthropic’s cloud payments could hit $6.4 billion in 2026 — money counted as revenue under one method and netted out under the other. If the SEC forces a harmonized approach, one company’s headline revenue could shrink by billions overnight.

💡 The takeaway: Accounting sounds boring — until it changes how a trillion-dollar IPO gets priced. This is the kind of detail that quietly decides who wins the public-market race.

⚡ QUICK HITS

• 25+ SpaceX ETFs Already Registered: Wall Street moved fast — over 25 SPCX-linked ETFs were registered before a single share traded, more than half of them leveraged. Fidelity even dropped its account minimum to $2,000 to widen retail access.

• MSCI Index Buying Kicks In: Because SpaceX listed with just a 4% float, MSCI-tracking passive funds must start buying SPCX today — a built-in wave of structural demand independent of the fundamentals.

• Anthropic Still Tops Business Adoption: The latest Ramp AI Index confirms Anthropic leads US business adoption at 34.4% vs OpenAI’s 32.3% — powered by Claude Code, now reportedly authoring 4% of all public GitHub commits worldwide.

• A More Cautious Read on Claude: A separate IDC survey of 1,000+ organizations found only 19% report “extensive” Claude use — trailing OpenAI and Google on depth-of-use, even as Anthropic wins new adoption.

• Bezos Warns of “Labor Scarcity”: Counterintuitively, Bezos predicts AI productivity gains will create a world where demand for human workers outpaces supply — even as Amazon automates aggressively.

• Blue Origin Setback: A New Glenn rocket exploded during a May 28 hot-fire test in Florida — a reminder that even as AI valuations soar, the hardware frontier remains unforgiving.

📊 AI NUMBER OF THE WEEK

$1.77 TRILLION

SpaceX’s market cap at the close of its first trading day — making it the largest IPO in recorded history, raising $75 billion and more than doubling the previous record set by Saudi Aramco. For context, that’s larger than the entire GDP of Italy. The AI infrastructure era now has its first true public-market titan.

👀 WHAT WE’RE WATCHING

The OpenAI and Anthropic listings.SpaceX just proved the market has an appetite for trillion-dollar AI debuts. Both OpenAI and Anthropic are expected to go public this fall, and SpaceX’s orderly ~25% pop removes a major excuse to delay. But with the Magnificent Seven already wobbling and an accounting dispute brewing, the next two listings will test just how much the public markets can absorb. We’ll be tracking every filing.