Nvidia and Quantum Machines, the Israeli startup offering an orchestration platform to controlling and operating quantum processors, today announced the launch of Nvidia DGX Quantum, which combines Nvidia’s Grace Hopper Superchip with Quantum Machines’ controller.

Over the course of the last few years, it’s become increasingly clear that to effectively operate a quantum computer, you’ll also need a lot of classical compute resources to control these systems. And these two systems also need to be tightly integrated, in part because you need to run calibration and error correction algorithms on the classical machines to keep the quantum machines running. But every additional compute cycle then also helps to run additional functions and evaluate the data from the quantum computer.

“Quantum computers do have a very natural place — and that is to power supercomputers,” said Quantum Machines co-founder and CEO Itamar Sivan. “The most natural place for them, I think, is in supercomputing infrastructure and cloud infrastructure. On the one hand, these will be the first places to integrate quantum computing at a greater scale. We want to power these supercomputers with quantum, but we also want to power the quantum with the supercomputers.”

Image Credits: Quantum Machines

That’s what this collaboration is all about. You can’t run a quantum computer without massive classical compute resources and for a supercomputer, quantum essentially becomes another co-processor for handling specific tasks, similar to how GPUs are often used to accelerate machine learning tasks, for example.

“If you have a million cubits or hundreds of thousands of cubits and each has a dozen parameters that need to be independently optimized — the leaders in the community are turning their attention towards AI methods, which shine on an NVIDIA platform,” explained Tim Costa, Nvidia’s director of high performance computing and quantum. “There’s this huge GPU compute requirement for standing up a computer which will deliver on the promise of quantum. And to do that, of course, we have to get tightly coupled to quantum, so we need to work with the leader in the field in terms of interfacing with quantum and controlling quantum, which is Quantum Machine.”

Nvidia teams up with Quantum Machines to combine classical and quantum machines by Frederic Lardinois originally published on TechCrunch

Nvidia announced some new features for its cloud gaming service during its virtual CES press conference. The company is upgrading its premium plan by adding new servers with better hardware components. Users on the $19.99 plan should expect better performance for more demanding games.

The company is now using GeForce RTX 4080-class graphics processing units on its high-end servers. Before today, users paying for the most expensive subscription plan could access servers with server-grade GPUs that are equivalent to GeForce RTX 3080 GPUs.

As a reminder, GeForce Now lets you play your own games from the cloud. The game is running in a data center near you and the video feed is then relayed to your device. GeForce Now supports Windows, macOS, Android (as well as Android TV) and some web browsers (including Safari on the iPhone and iPad).

GeForce Now customers still have to buy games on Steam, the Epic Games Store and other digital stores — they own the games even if they stop subscribing to the service. But the biggest issue with the service is that some game publishers refuse to let Nvidia support their games on GeForce Now. There are currently 1,500 supported games, including Fortnite, League of Legends, Cyberpunk 2077 and many Ubisoft games. But you can’t play Overwatch 2 or Elden Ring for instance.

Customers can try out GeForce Now for free. There is a queue system and you are limited to 60-minute gaming sessions. If you want to use the service on a daily basis, a ‘Priority’ membership lets you launch a game right away and play for up to six hours at a time for $9.99 per month. You are limited to a 1080p resolution and 60 frames per second.

Last year, Nvidia added a premium tier called GeForce Now RTX 3080 for $19.99 per month. Because of today’s update, this tier is getting a new name. The company is now calling it GeForce Now Ultimate.

In addition to access to more powerful servers, GeForce Now Ultimate supports 4K resolution. If you have a gaming monitor, the Ultimate membership now also supports 240Hz (up from 120Hz). Users can also enable Nvidia’s proprietary features, such as DLSS 3 and Nvidia Reflex.

If you have an Nvidia G-Sync monitor, GeForce Now will adapt the streaming rate depending on how many frames per second you get in your Nvidia Reflex-compatible game. That’s neat! But if you have an Nvidia G-Sync monitor, you likely also have a gaming PC so you may not need GeForce Now.

Existing GeForce Now RTX 3080 members are going to be automatically upgraded to the GeForce Now Ultimate plan in late January. GeForce Now Ultimate will still cost $19.99 per month.

Nvidia upgrades GeForce Now with RTX 4080 performance for premium users by Romain Dillet originally published on TechCrunch

Run.ai, the well-funded service for orchestrating AI workloads, made a name for itself in the last couple of years by helping its users get the most out of their GPU resources on-premises and in the cloud to train their models. But it’s no secret that training models is one thing while putting them into production is another — and that’s where a lot of these projects still fail. It’s maybe no surprise then that the company, which sees itself as an end-to-end platform, is now moving beyond training to also support its customers in running their inferencing workloads as efficiently as possible, whether that’s in a private or public cloud, or on the edge. With this, the company’s platform now also offers an integration with Nvidia’s Triton Inference Server software, thanks to a close partnership between the two companies.

“One of the things that we that we identified in the last 6 to 12 months is that organizations are starting to move from building and training machine learning models to actually having those models in production,” Run.ai co-founder and CEO Omri Geller told me. “We started to invest a lot of resources internally in order to crack this challenge as well. We believe that we cracked the training part and built the right resource management there, so we are now focused now on helping organizations manage their compute resources for inferencing, as well.”

Image Credits: Nvidia

The idea here is to make it as easy as possible for businesses to deploy their models. Run.ai promises a two-step deployment process that doesn’t involve writing YAML files. Thanks to Run.ai’s early bet on containers and Kubernetes, it is now able to move these inferencing workloads on the most efficient hardware and with the new Nvidia integration into the Run.ai Atlas platform, users can even deploy multiple models — or instances of the same model — on the Triton Inference Server, with Run.ai, which is also part of Nvidia’s LaunchPad program, handling the auto-scaling and prioritization on a per-model basis.

While inferencing doesn’t require the same kinds of massive compute resources it takes to train a model, Nvidia’s Manuvir Das, the company’s VP of enterprise computing, noted that these models are becoming increasingly large and deploying those on a CPU just isn’t possible. “We built this thing called Triton Inference Server, which is about doing your inference not just on CPUs but also on GPUs — because the power of the GPU has begun to matter for the inference,” he explained. “It used to be you needed the GPU to do the training and once you’ve got the models, you could happily deploy them on CPUs. But more and more, the models have become bigger and more complex. So you need to actually run them on the GPU.”

And as Geller added, models will only get more complex over time. He noted that there is, after all, a direct correlation between the computational complexity of models and their accuracy — and hence the problems businesses can solve with those models.

Even though Run.ai’s early focus was on training, the company was able to take a lot of the technologies it built for that and apply them to inferencing as well. The resource-sharing systems the company built for training, for example, also apply to inferencing, where certain models may need more resources to be able to run in real time.

Now, you may think that these are capabilities that Nvidia, too, could maybe built into its Triton Inference Server, but Das noted that this is not the way the company is approaching the market. “Anybody doing data science at scale needs a really good end-to-end ML ops platform to do the whole thing,” he said. “That’s what Run.ai does well. And then what we do underneath, we provide the low-level constructs to really utilize the GPU individually really well and then if we integrate it right, you get the best of both things. That’s one reason why we worked well together because the separation of responsibilities has been clear to both of us from the beginning.”

It’s worth noting that in addition to the Nvidia partnership, Run.ai also today announced a number of other updates to its platform. These include new inference-focused metrics and dashboards, as well as the ability to deploy models on fractional GPUs and auto-scaling them based on their individual latency Service Level Agreements. The platform can now also scale deployments all the way to zero — and hence reduce cost.

One more hurdle up ahead for Activision Blizzard, the games giant behind “Call of Duty” that Microsoft is looking to acquire for $68.7 billion. The UK’s Competition and Markets Authority has announced a formal investigation into the proposed deal. This opens the investigation up for feedback from “any interested party” ahead of the CMA deciding whether to embark on a phase-2, deeper enquiry into whether the deal is anticompetitive and represents antitrust violations in the U.K. Those interested parties have until September 1 to respond.

Areas that it will be assessing include whether the deal leads to higher prices, lower quality, or reduced choice in games and the gaming ecosystem.

“The Competition and Markets Authority (CMA) is considering whether it is or may be the case that this transaction, if carried into effect, will result in the creation of a relevant merger situation under the merger provisions of the Enterprise Act 2002 and, if so, whether the creation of that situation may be expected to result in a substantial lessening of competition within any market or markets in the United Kingdom for goods or services,” it notes.

If the acquisition goes ahead, it will be one of the biggest M&A deals ever in technology, led by one of the world’s biggest tech companies and with some of the world’s most popular brands in digital entertainment — which in addition to “Call of Duty” also include “World of Warcraft” and “Candy Crush.” So in that regard the antitrust investigation is a fairly routine move. As the CMA notes, “The deal is set to be reviewed by competition authorities around the world and, as is usual practice, the CMA will engage with its counterparts as appropriate.”

The CMA notes that its more formal criteria for assessing investigations is that it can do so in cases where two or more enterprises cease to be distinct and “either the UK turnover of the acquired business exceed £70 million, or the 2 businesses supply/acquire at least 25% of the same goods/services supplied in the UK and the merger increases the share of supply.”

The U.S. has a similar set of rules in place for triggering antitrust enquiries and so unsurprisingly, the FTC in the U.S. is also currently investigating the deal. The regulators have been known to scupper, or add provisions, to deals, as well as nod them through.

The CMA has played a significant role in recent years in the fate of several large tech companies’ business development strategies. It struck down the acquisition of Arm by Nvidia, but it nodded through Microsoft’s acquisition of Nuance for $20 billion. It’s currently also investigating Google’s adtech stack and at long last is looking into the duopoly in mobile in the country that is Apple and Google. Although it didn’t raise much noise in Facebook’s acquisition of WhatsApp several years ago, more recently it ordered the company to sell Giphy. One route to getting past regulatory concerns at the CMA has seen companies downsize their acquisitions, such as eBay and Adevinta did when selling off several assets to get their classifieds deal past them. 

The news of the antitrust probe comes at a time when Activision Blizzard has already weathered a number of controversies particularly around its labor and wider human relations practices.

When Microsoft’s bid for Activision was first announced in January of this year, quality assurance testers at Raven Software, a division of Activision, had been on strike for about five weeks over the termination of contracts for some workers and what they saw as unfair treatment by the company, given the stress that this group faces in their daily work. They voted to unionize in May, making it the first union in a major gaming company.

Beyond that, the company has been under intense public and regulatory scrutiny over workplace culture. The company, which overall employs about 10,000 people globally, was the subject of a two-year investigation by the California Department of Fair Employment and Housing. It eventually filed a lawsuit against Activision Blizzard in July 2021, claiming a “‘frat boy’ workplace culture” at the company and describing it as a “breeding ground for harassment and discrimination against women.” On top of this, Bobby Kotick, who has been the CEO of the company since 1991 (originally at Activision pre- its Blizzard merger), reportedly knew about, yet failed to act, over sexual misconduct and rape allegations.

To be clear, details of the HR drama at the company are not within the parameters of the antitrust investigation, but they do contribute to the bigger picture of a company under the gun.

Those who are giving feedback on the merger to the CMA can do so here and have until September 1 to make their submissions.

Strong Compute, a Sydney, Australia-based startup that helps developers remove the bottlenecks in their machine learning training pipelines, today announced that it has raised a $7.8 million seed round. The round includes a total of 30 funds and angels, including the likes of Sequoia Capital India, Blackbird, Folklore and Skip Capital, as well as Y Combinator, Starburst Ventures and founders and engineers from companies like Cruise, Waymo, Open AI, SpaceX and Virgin Galactic.

The company, which was part of Y Combinator’s Winter ’22 batch, promises that its optimizations can speed up the training process by 10x to 1000x, depending on the model, pipeline and framework. As Strong Compute founder Ben Sands, who previously also co-founded AR company Meta, told me, the team has recently made some breakthroughs where it was able to take Nvidia’s reference implementation, which its customer LayerJot used, to run 20 times faster.

Image Credits: Strong Compute

“That was a big win,” Sands said. “It really gave us the sense that there is nothing that can’t be improved.” He didn’t quite want to reveal all of the details of how the team’s optimizations worked, but he noted that the company is now hiring mathematicians and is building tools that give it a more detailed view of how their user’s code interacts with the CPUs and GPUs at a much deeper level than was previously possible.

As Sands stressed, the current focus for the company is to start automating a lot of the current work to optimize the training process — and that’s something the company can now tackle, thanks to this funding round. “Our goal now is to have serious development partners in self-driving, medical and aerial, in order to be looking at what is actually going to generalize really well,” he explained. “We’ve now got the resources to have an R&D team that doesn’t have to deliver something in a two-week sprint but that can actually look at what’s some real core tech that could take a year to actually get a win out of but that can really help with that automated analysis of the problem.”

The company currently has six full-time engineers but Sands plans to double that over the next few months. In part, that’s also because the company is now getting inbound interest from large companies that often spend $50 million or more on their compute resources (and Sands noted that the market is basically bi-modal, with customers either spending less than $1 million or $10 to $100 million, with only a few players in the middle).

Image Credits: Strong compute

Every company that is trying to build ML models, though, suffers from the same problem: training models and running experiments to improve them still take a lot of time. That means the well-paid data scientists working on these problems spend a lot of time in a holding pattern, waiting for results to come in. “Strong Compute is solving the basketball court problem,” said SteadyMD CFO Nikhil Abraham. “Long training times had our best devs shooting hoops all day, waiting on machines.”

And while some of that inbound interest is coming from the financial industry and companies that want to optimize their natural language processing models, Strong Compute’s focus remains on computer vision for the time being.

“We’ve only just scratched the surface of what machine learning and AI can do.” said Folklore partner Tanisha Banaszcyk. “We love working with founders who have long-range ambition and visions that will endure across generations. Having invested in autonomous driving, we know how important speed to market is – and see the impact Strong Compute can have on this market with its purpose-built platform, deep understanding of the $500B market and world-class team.”

Vultr, the cloud platform that specializes in providing access to basic infrastructure services at a relatively low cost, today announced the launch of Vultr Talon, a new service that will offer developers access to virtualized GPUs, starting with Nvidia’s A100 Tensor Core GPU, with prices as low as $0.134/hour (or $90/month) for access to a twentieth of the compute power of an A100.

Typically, when you need access to a high-end GPU for machine learning and similar use cases, the smallest unit you can buy is access to the entire GPU. On the CPU side of things, buying just a slice of CPU power is standard, but until now, that just wasn’t possible for GPUs, even though for many use cases, like inferencing, you would only need access to a fraction of the GPU’s compute power.

Image Credits: Vultr

Vultr argues that its platform is the first to offer fractional access to these high-end GPUs. The service is powered by the Nvidia AI Enterprise software suite and Vultr CEO J.J. Kardwell, who joined the company in late 2020, believes that this is another point in his service’s favor. “If you think about the way that the big clouds deploy, they want to buy GPU hardware and then push their own stack, whereas we partnered very closely with Nvidia to deliver not just the GPU itself but the Nvidia AI Enterprise software stack and the full set of libraries,” he said. “The best way to get the most out of the GPU — the physical GPUs — is through that Nvidia software stack — and that’s a very, very different approach.”

Image Credits: Vultr

The company, which also recently launched its managed Kubernetes platform into general availability, always remained a bit under the media radar — and much of that was by design. It never raised a lot of outside funding since its launch in 2014 and mostly relied on word of mouth to grow to $125 million ARR today, as Kardwell told me. That’s the way the company’s founder David Aninowsky, who is now its executive chairman, wanted to build Vultr. But as the company is now reaching this scale — and likely thinking about an IPO at some point in the future — the team is clearly looking to elevate its profile a bit more.

“We really believe we’re doing more than anyone to democratize access to cloud computing globally. And as we seek to do that, we think it’s important that people understand that mission,” he explained.

The A100 is now available in Vultr’s New Jersey data center, with other locations following in the next few weeks. The company also tells me that it will add other Nvidia GPUs to its lineup later this year.

OpenStack, the massive open source infrastructure-as-a-service project that allows enterprises and public hosting services to run their own on-premise clouds, today released version 25 of its software, dubbed Yoga. Like all large open source projects, OpenStack has gone through its ups and downs, but as the Open Infrastructure Foundation, the organization behind OpenStack and a number of other projects like Kata Containers, recently announced, OpenStack now manages over 25 million CPU cores in production and nine out of the top 10 telcos now run OpenStack, as do large enterprises like Bloomberg, Walmart, Workday and TechCrunch parent Yahoo. China Mobile alone runs an OpenStack deployment with 6 million cores, and more than 180 public cloud data centers now run OpenStack.

Even after 25 releases in 12 years, the OpenStack community is still adding new features, in addition to the usual bug fixes and maintenance updates. As with most recent releases, this means increased support for additional hardware, for example. With Nvidia now a major contributor to OpenStack, there is new support for SmartNIC DPUs, that is, the ability to offload network processing to specialized cards, in OpenStack’s core networking and compute services. The OpenStack Cinder storage service now also supports LightOS for new storage types like NVMe/TCP and NEC V Series Storage. Not something you’d need for a small deployment, but features that will matter to some of OpenStack’s largest users.

“It’s great to see all of those hardware manufacturers getting involved directly in OpenStack to make sure that we correctly support and expose the features in their hardware,” said the Open Infrastructure Foundation general manager Thierry Carrez. It’s worth noting that the general manager position is still quite new, with Carrez moving to the job in January of this year after being the VP of Engineering for the OpenStack/Open Infrastructure Foundation for many years. In this new role, he now oversees the foundation’s operations, covering engineering, product, community and marketing.

Other major updates in this release include a new soft delete scheme for OpenStack Manila, the project’s shared file system service. Kendall Nelson, the senior upstream developer advocate for the Open Infrastructure Foundation, likened this to the recycle bin on your desktop. “It’s one of those things where it’s like, you know, why don’t we do that? We could have been doing this the whole time and I think that Manila has been pretty stable for a while, so [the developers were] like, ‘Oh, well, let’s go and do the obvious things that we could have done all along’,” she said.

With this release, OpenStack is also expanding its support for a number of cloud-native infrastructure projects like the popular Prometheus monitoring system and Kuryr and Tacker Kubernetes tools, and welcoming two new projects, the Skyline dashboard and Venus log management module.

As for the Foundation itself, it’s worth noting that 12 new companies recently joined the organization. These new members are mostly in the lower silver tier, like B1 Systems, Okestro, OpenMetal and TransIP, but Vexxhost, for example, is joining as a gold member. Overall, the organization’s corporate membership is up 20% since November 2021.

Later this year, the Open Infrastructure Foundation will also host its first in-person conference again, the OpenInfra Summit in Berlin, Germany, in early June. “When we launched the Open Infra Foundation, we said we were going to bring together a community to build the next decade of open infrastructure after 10 years of OpenStack and related projects,” Open Infrastructure Foundation CEO and Executive Director Jonathan Bryce said. “We’re a year into it and it’s been really exciting to see this coalition of companies who are joining across vendors, new tech leaders like Nvidia, new users like BBC and others, along with the projects that are coming in. I’m really excited to finally get all of these people back together for the first time since the pandemic.”

Soon, the OpenStack project will also change its release cadence. Currently, the community is publishing two releases a year. Starting in 2023, it’s moving to a “tick-tock” schedule, with one major and one minor release on the same six-month cadence as today. In part, this is because of feedback from operators who don’t want to have to upgrade their environments every six months.

“This really helps the smaller OpenStack clouds, because you can reliably like ‘okay, well, we have a year to breathe now before the next one,’ as opposed to every six months,” Bryce said. “There’s much less risk than there used to be in terms of upgrading, but it’s still a lot of work, so for those companies that use OpenStack that are a little bit smaller and have smaller teams or maybe are newer to it, this like long-term support cadence should really help them get off the ground and get moving. And then, the bigger companies that are used to the six-month release cadence like Red Hat are still going to be able to get their features right when they’re coming out.”

At its annual GTC conference for AI developers, Nvidia today announced its next-gen Hopper GPU architecture and the Hopper H100 GPU, as well as a new data center chip that combines the GPU with a high-performance CPU, which Nvidia calls the “Grace CPU Superchip” (not to be confused with the Grace Hopper Superchip).

The H100 GPU

With Hopper, Nvidia is launching a number of new and updated technologies, but for AI developers, the most important one may just be the architecture’s focus on transformer models, which have become the machine learning technique de rigueur for many use cases and which powers models like GPT-3 and asBERT. The new Transformer Engine in the H100 chip promises to speed up model training by up to six times and because this new architecture also features Nvidia’s new NVLink Switch system for connecting multiple nodes, large server clusters powered by these chips will be able to scale up to support massive networks with less overhead.

“The largest AI models can require months to train on today’s computing platforms,” Nvidia’s Dave Salvator writes in today’s announcement. “That’s too slow for businesses. AI, high performance computing and data analytics are growing in complexity with some models, like large language ones, reaching trillions of parameters. The NVIDIA Hopper architecture is built from the ground up to accelerate these next-generation AI workloads with massive compute power and fast memory to handle growing networks and datasets.”

The new Transformer Engine uses customer Tensor Cores that can mix 8-bit precision and 16-bit half-precision as needed while maintaining accuracy.

Image Credits: Nvidia

“The challenge for models is to intelligently manage the precision to maintain accuracy while gaining the performance of smaller, faster numerical formats,” Salvatore explains. “Transformer Engine enables this with custom, NVIDIA-tuned heuristics that dynamically choose between FP8 and FP16 calculations and automatically handle re-casting and scaling between these precisions in each layer.”

The H100 GPU will feature 80 billion transistors and will be built using TSMC’s 4nm process. It promises speed-ups between 1.5 and 6 times compared to the Ampere A100 data center GPU that launched in 2020 and used TSMC’s 7nm process.

In addition to the Transformer Engine, the GPU will also feature a new confidential computing component.

Grace (Hopper) Superchips

Grace Superchip

The Grace CPU Superchip is Nvidia’s first foray into a dedicated data center CPU. The Arm Neoverse-based chip will feature a whopping 144-cores with 1 terabyte per second of memory bandwidth. It actually combines two Grace CPUs connected over the company’s NVLink interconnect — which is reminiscent of the architecture of Apple’s M1 Ultra.

The new CPU, which will use fast LPDDR5X memory, will be available in the first half of 2023 and promises to offer 2x the performance of traditional servers. Nvidia estimates the chip will reach 740 points on the SPECrate®2017_int_base benchmark, which would indeed put it in direct competition with high-end AMD and Intel data center processors (though some of those score higher, but at the cost of lower performance per watt).

“A new type of data center has emerged — AI factories that process and refine mountains of data to produce intelligence,” said Jensen Huang, founder and CEO of Nvidia. “The Grace CPU Superchip offers the highest performance, memory bandwidth and NVIDIA software platforms in one chip and will shine as the CPU of the world’s AI infrastructure.”

In many ways, this new chip is the natural evolution of the Grace Hopper Superchip and Grace CPU the company announced last year (yes, these names are confusing, especially because Nvidia called the Grace Hopper Superchip the Nvidia Grace last year). The Grace Hopper Superchip combines a CPU and GPU into a single system-on-a-chip. This system, which will also launch in the first half of 2023, will feature a 600GB memory GPU for large models and Nvidia promises that the memory bandwidth will be 30x higher compared to a GPU in a traditional server. These chips, Nvidia says, are meant for “giant-scale” AI and high-performance computing.

The Grace CPU Superchip is based on the Arm v9 architecture and can be configured as standalone CPU systems or for servers with up to eight Hopper-based GPUs.

The company says it is working with “leading HPC, supercomputing, hyperscale and cloud customers,” so chances are these systems are coming to a cloud provider near you sometime next year.

Gremlin, the popular chaos engineering startup, today announced that it has brought on Josh Leslie, the former CEO of Cumulus Networks, which was acquired by Nvidia in 2020, as its CEO. Kolton Andrus, who co-founded the company back in 2016 and held the CEO role since, will become Gremlin’s CTO to focus on driving the company’s product forward.

In conjunction with this leadership change, Gremlin also today announced that it passed $10M in annualized recurring revenue in 2021 and now has a customer base that includes the likes of GrubHub, JP Morgan Chase, Target and Twilio. The company expects to see 100% growth in 2022.

In many ways, this is very much the story of a technical founder wanting to get back to his roots and do what he does best, with an experienced CEO stepping in to guide the company through its next phase.

“I’m an engineer by trade. I built chaos engineering tooling for Amazon and Netflix and that was what really led to the foundation of the company,” Andrus told me. “At that point, it’s ‘let’s go figure out if there’s a market here, let’s go figure out if this product is going to work, let’s go figure out if we can build a meaningful business out of this?'”

Having proven all of those points and having become close to the de facto standard for chaos engineering tools, Andrus and the team looked at what was needed to take the company further. “I feel like I’ve accomplished a good amount for that first step, but I [have to] look at what is it going to take for us to go from a $10 million company to a $50 or $100 million company — and how I can best use my time to help impact that,” Andrus explained.

Image Credits: Corey Thomas/Gremlin

Leslie, after taking some time off after the sale of Cumulus, joined the Gremlin board last year. “When we sold cumulus, I had the opportunity to take a little break from full-time work and get involved with a broader set of companies,” he told me. “I had a chance to advise and invest in a whole set of companies, one of which was Gremlin. I got to get to know Kolton and get to understand the business and get to understand the market — and I got to do that over a pretty long period of time. That’s really just a privilege and really good fortune to have the opportunity to really understand the business over a period of time, but more importantly, understand the relationships. Having done the CEO job once, I think one thing I can take away is it’s all about the team, It’s all about the relationships, it’s all about the trust.”

For the next phase of Gremlin, the company is looking to invest in its sales and marketing team, but also to build the product features that can help it take chaos engineering a bit more mainstream. Leslie actually believes that the market opportunity for Gremlin is significantly larger than Cumulus’ ever was, given that basically every company that has even a semi-complex infrastructure stack could be a customer — and that’s pretty much every company, given the cost involved in having their infrastructure fail.

“A lot of what we’ve done in the last five years is building the category — going out, teaching people what it means and how to do it and how to do it well, whether they are customers or not,” Andrus said. “We’re going to continue on in that approach because building a large category serves us well but it also serves everybody well.”

Nvidia’s deal to acquire Arm is off, the two companies and Arm’s owner SoftBank announced tonight. SoftBank and Nvidia today agreed to drop their efforts to acquire Arm. With this, there is also a major leadership change at Arm. The company’s current CEO, Simon Segars, is leaving his post, effective today, with Rene Haas, the president of Arm’s IP group (and former Nvidia VP and general manager of its computing products business), taking his role as the company says it is exploring a public offering in lieu of the acquisition. The current plan is for the IPO to happen sometime within the next 12 months.

The Financial Times first reported that the deal had collapsed earlier today.

As Haas told me in an interview shortly before today’s announcement, Segars decision to leave the company was very much a personal choice. “He has decided that at this stage of his career, the time and energy required to take the company public and everything around that was not something he wanted to sign up to,” Haas said. “So he’s going to step down. I’m going to take over for him.”

“Rene is the right leader to accelerate Arm’s growth as the company looks to re-enter the public
markets,” said Masayoshi Son, Representative Director, Corporate Officer, Chairman and CEO of SoftBank Group Corp. “I would like to thank Simon for his leadership, contributions and dedication to Arm over the past 30 years.”

Haas noted that Arm’s business today is stronger than ever. “We are very enthusiastic and excited about this next chapter for the company,” he told me. “The company’s been doing great. […] Revenue and profitability levels never seen by Arm before and certainly not seen before we joined SoftBank. But, probably more importantly, the diversification of the business relative to what it looked like pre-SoftBank, we’re a much stronger company now in areas like the cloud and the data center, we’re a much stronger company and markets like automotive and we have a huge opportunity for future markets such as IoT and metaverse.”

While Haas wouldn’t comment any further on Nvidia’s and SoftBank’s decision, he noted that the deal finally collapsed today, but he admitted that the discussions about a leadership change had started earlier.

“While we are disappointed that the acquisition did go through, we are at the same time, very excited about our prospects going forward and can’t wait to get this next chapter started,” he said. What exactly that’ll look like, Haas wouldn’t say just yet, but he noted that the company will continue its efforts to push into the CPU and GPU market, as well as continue its efforts in the AI space. “Continuing what we’ve been doing and executing on that is going to be really important, because we’ve demonstrated a recipe on how to grow the business and we definitely want to continue that,” he said.

In some ways, today’s announcement doesn’t come as a surprise. After the U.S. Federal Trade Commission announced that it would sue to block the merger, arguing that the combined company would be able to “unfairly undermine Nvidia’s rivals,” Bloomberg reported that Nvidia was preparing to abandon its plans. In the U.K., where Arm is headquartered, the merger hit similar roadblocks in recent months, as well as with EU antitrust regulators. In the end, an Nvidia that dominated the GPU and AI accelerator market and owned the IP to the chips that power virtually every smartphone and IoT device rang all kinds of alarm bells. In the end, the two companies would’ve likely had to make considerable changes to their deal to get it through the regulatory process — or abandon it.

Nvidia first announced its plans for this mega-merger in September 2020. At the time, Nvidia CEO Jensen Huang argued that this would allow his company to create “a company fabulously positioned for the age of AI.”

Throughout all of this, Nvidia’s and Arm’s leadership publicly remained optimistic that the deal would eventually pass muster with the regulatory authorities. Knowing that there would be some pushback, the two companies had always given themselves a lot of time to close this deal, with an expected closing date of March 2022, 18 months after the announcement. In recent months, both companies admitted that they would miss this date. They were also up against a bit of a deadline of September 2022, after which SoftBank would keep its $1.25 billion breakup fee.

News reports surfaced over the past 24 hours that the $40 billion Nvidia-Arm deal, which ranks among the most expensive tech deals ever, is in peril. Nvidia is reportedly ready to walk away due to regulatory pressure. The question is, what does it mean for tech M&A if this deal falls apart?

Let’s not forget that last year at this time Visa shut down a $5.3 billion deal to acquire Plaid after the U.S. Justice Department gave it a closer look than made the credit card giant comfortable. Just last month, the U.K.’s antitrust watchdog announced it was holding up Microsoft’s proposed $20 billion acquisition of Nuance Communications. That deal remains in limbo while it decides what to do with it, and there is also a possibility that the country’s Competition and Markets Authority (CMA) will open an investigation as well.

It’s worth noting that EU authorities cleared the deal last month without conditions.

Now we have Nvidia facing much broader regulatory scrutiny as international regulators worry about the combined companies shifting the competitive balance in the chip market.

Geoff Blaber, Chief Executive Officer at analyst firm CCS Insight says that this deal faced tough regulatory headwinds since it was announced, and it’s not surprising to him that Nvidia would decide to walk away.

“The Nvidia–Arm deal has faced intense scrutiny and pressure from the start and it’s no surprise the deal is in danger of collapse. Finding a way to appease regulators whilst maintaining the value and justifying the $40 billion price tag has proven overwhelmingly challenging,” Blaber said.

He added that the company could try an alternate exit, but it won’t provide the same rate of return for investors that the Nvidia deal would have. “It has also proven disruptive to Arm and its ecosystem in the process. An IPO is an alternative path, but is unlikely to provide Softbank (Arm’s primary investor) a comparable return.”

Patrick Moorhead, founder and principal analyst at Moor Insight & Strategies agrees that it puts Arm in a more difficult financial position, but he sees Nvidia coming out pretty much unscathed, even if it was not able to get the company it wanted.

“For Arm, it means an IPO and a slightly weaker company without Nvidia’s capitalization. For Nvidia, it’s business as usual. Nvidia gets an architectural license if the deal falls apart which means it can, for no license fee, create its own custom CPUs,” putting the company in good shape no matter what happens in this deal.

That could be a big part of why Nvidia with so much regulatory scrutiny simply decided it was no longer worth the effort, especially since it could essentially have its cake and eat it too and it could put that $40 billion into other areas of investment to drive growth in the future.

It could be that this is a unique situation and that it doesn’t really have much impact on the broader M&A landscape, but as we see more careful oversight of deals, and the on-going antitrust efforts in the U.S. involving big tech, it certainly feels like there could be more here than one company growing tired of a bureaucratic process.

There has been talk of governments in general looking at tech deals more closely than in the past, but with the EU all but rubber stamping the Microsoft-Nuance deal, it could depend on the mechanics of each deal, the companies involved, and especially the perceived impact on competitive balance.

NVIDIA has reportedly made little to no progress in gaining regulatory approval for its $40 billion purchase of ARM and is privately preparing to abandon the deal, according to Bloomberg‘s sources. Meanwhile, current ARM owner SoftBank is reportedly advancing a program to take ARM public as an alternative to the acquisition, said another person familiar with the matter.

NVIDIA announced the deal in September 2020, with CEO Jensen Huang proclaiming it would “create a company fabulously positioned for the age of AI.” ARM’s designs are used under license almost universally in smartphones and other mobile devices by companies like Apple, Qualcomm, Microsoft, Samsung, Intel and Amazon.

A backlash began soon after the announcement. The UK, where ARM is based, launched an antitrust investigation into the acquisition in January 2021, and another security probe last November. In the US, the FTC recently sued to block the purchase over concerns it would “stifle” competition in industries like data centers and car manufacturing. China would also reportedly block the transaction if other regulators don’t, Bloomberg‘s sources say.

We continue to hold the views expressed in detail in our latest regulatory filings — that this transaction provides an opportunity to accelerate Arm and boost competition and innovation.

Companies like Intel, Amazon and Microsoft have reportedly given regulators enough information to kill the deal, the sources say. They previously argued that NVIDIA can’t preserve ARM’s independence because it’s an ARM client itself. As such, it could also potentially become both a supplier and competitor to ARM licensees.

Despite the stiff headwinds, both companies maintain that they’re still pushing forward. “We continue to hold the views… that this transaction provides an opportunity to accelerate ARM and boost competition and innovation,” NVIDIA spokesman Bob Sherbin told Bloomberg. “We remain hopeful that the transaction will be approved,” a SoftBank spokesperson added in a statement.

Despite the latter comment, factions at Softbank are reportedly pushing for an ARM IPO as an alternative to the acquisition, particularly while the semiconductor industry is so hot. Others in the company want to continue pursuing the transaction given that NVIDIA’s stock price has nearly doubled since it was announced, effectively increasing the transaction price.

The initial agreement expires on September 13th, 2022, but will automatically renew if approvals take longer. NVIDIA predicted that the transaction would close in approximately 18 months — a deadline that now seems unrealistic.

Editor’s note: This article originally appeared on Engadget.