AI lab Mirendil has signed a multi-year partnership with Google Cloud to source compute capacity for its self-improving AI research, TechCrunch has exclusively learned. 

The deal mirrors two trends shaping the AI industry: cloud giants are courting startups with huge infrastructure commitments, and AI companies are snatching up as many compute deals as they can to secure access as they scale. 

The deal is worth upwards of $100 million, Mirendil’s co-founder and CEO, Benham Neyshabur, told TechCrunch. That’s roughly half of what Mirendil raised in seed funding at a $1 billion valuation in late June.

The deal gives the startup access to both Google’s TPUs and Nvidia GPUs, as well as managed training clusters with which Mirendil will work on its self-improving AI. The startup hopes its AI will eventually be able to take on the work of an entire frontier AI lab.

Self-improving AI, also known as recursive self-improvement, refers to AI systems that iteratively improve themselves. It’s a concept that major labs like Anthropic, where Mirendil’s co-founders hail from, have been working on. A handful of startups like Recursive Superintelligence and Ricursive Intelligence have also recently sprung up around achieving that goal.

Mirendil believes this process will automate a lot of scientific and AI research, helping scientists make progress in fields like medicine, biology, and materials science. 

Neyshabur thinks AI can mimic how human scientists can learn more about new domains, accumulate knowledge and expertise, and gradually improve their performance. “You can have a self-improving AI where you can point a problem at it and it keeps getting better with time,” he said.

“How can we have an AI system that keeps doing research, keeps improving its own knowledge and performance when it comes to Alzheimer’s disease?” he continued. “This technology allows us to set goals that are ambitious for AI, and the AI would keep making progress.”

Training self-improving AI, however, requires enormous amounts of computing power. The lab’s co-founder Harsh Mehta said training is increasingly about matching the right workloads to the right hardware. 

“These models are really good at working with different workloads and chips, and assigning the right workloads to the right chips,” Mehta said. “[Google] provides multiple kinds of chips […] this flexibility allows us to ultimately mix and match workloads with the right kind of accelerators, and then lower the cost not just for us, but also for our customers using our systems.”

That flexibility is central to Google’s AI infrastructure pitch. Amin Vahdat, SVP and chief technologist of AI and infrastructure at Google, said in a statement that AI advancement isn’t just about chip-level performance anymore, “but how we orchestrate entire systems of intelligence and break through the physical constraints of scaling.”

Neyshabur said Mirendil’s software and systems layer help customers get more out of Google’s hardware, giving the cloud giant another potential leg up in the race against its competition. In return, Google gets a strategic partner building frontier recursive self-improving AI — technology that it can eventually shop around to enterprise customers. 

Ford’s next-generation electric truck will be called Fathom, and it will start at $28,350 when it goes on sale in 2027, the company said on Thursday.

A “destination and delivery” fee will bring that price up to $29,945, putting the Fathom just below the $30,000 mark that Ford has targeted since it first started talking more openly about the project in 2025.

The truck, when it arrives next year, represents a major bet for Ford. It’s the automaker’s second all-electric truck after the F-150 Lightning (third, if you count the limited-run Ranger EV from the late 1990s). But like many of its peers, the company has dramatically pulled back on its once-ambitious plans to build electric vehicles.

The Fathom is the first vehicle Ford is building on its new “Universal EV Platform,” or UEV, which will power a number of different EVs and involves a brand-new assembly line process meant to cut costs. It’s an attempt by CEO Jim Farley to try and catch up to category leader Tesla, as well as the many Chinese EVs that he’s argued against allowing into the U.S. market — all while building the truck in Louisville, Kentucky.

Deliveries begin in “fall 2027,” and Ford said on Thursday that it won’t reveal the truck until “early 2027,” which is when it will start taking pre-orders.

We do have some details, though. Ford has already said the Fathom will have Apple Maps built into its operating system, and on Thursday, the company said the truck will support Android Auto and Apple CarPlay. Notably, it will not support Apple’s next-generation CarPlay Ultra.

The Fathom will have five seats, which is more interior space than a Toyota Rav4; a “large, high-resolution touch screen,” and a “digital car key.” It will be capable of bidirectional power, much like the Lightning and the hybrid F-150.

Every Fathom will also be “BlueCruise-capable,” meaning it will come with the tech integrated, but buyers will likely have to pay extra to enable Ford’s hands-free advanced driver assistance system. The truck is supposed to have a newer, more powerful version of BlueCruise that Ford says will eventually be capable of handling supervised driving from starting point to destination. Eyes-off driving is expected in 2028.

The Fathom’s starting price also only includes the “standard range” lithium iron phosphate battery pack. Ford hasn’t yet said how many miles the truck can drive on a single charge.

Ford said the name represents “a deep understanding of the person who would drive” the truck, and how their “motivations look a little different from the traditional truck customer’s.”

“[T]he name had to do more than sound good. It had to capture a mindset, shifting the focus from specs to a spirit of adaptability,” the company wrote.

Electric trucks have not done terribly well in the U.S.. The Lightning only moved a little more than 10,000 units in its best quarter, and though Tesla’s Cybertruck had a higher peak, it has since crashed out.

Ford is gambling by choosing the truck form factor for its first UEV platform vehicle. But there are some inherent upsides: The Fathom and other vehicles built on the platform will be the first that Ford has designed from the ground up to use electric powertrains. That should mean fewer tradeoffs on range, efficiency, and overall packaging of the vehicle.

Both the Fathom and UEV have been in the works for a few years. It started as a skunkworks project that Farley tasked former Tesla exec Alan Clarke with leading, as TechCrunch first reported in 2024. The team is flush with talent from Rivian, Tesla and Apple’s now-defunct car project, and has purportedly been working in isolation in Los Angeles, separated from the bureaucracy of the rest of the company.

All of this likely contributed to getting the Fathom’s starting price to just under $30,000, a major accomplishment in a market where the average new car costs $50,000. The truck won’t have a lot of competition in that range — it’s essentially going up against just the spartan, Jeff Bezos-backed Slate pickup, and a few non-truck EVs like the Chevy Bolt.

Ford has found a ton of success with its small, low-cost Maverick pickup, which comes in both internal combustion and hybrid powertrain variants. The Fathom is an attempt to bottle that lightning, without going the way of the Lightning.

The customer support industry has seen a massive influx of startups like Sierra, Decagon and Parloa trying to infuse AI into automating customer calls, chat and messages so companies can handle more queries at scale.

But according to Athens-based Omilia, which has been working on automating voice calls and customer support since 2002, throwing AI at every process can be wasteful. The company’s CEO Dimitris Vassos says that a large proportion of incoming customer support queries involve basic information, such as account balances, and there’s little need to deploy large language models for such tasks.

“We will use any available weapon to win the battle for customer service. Companies like Sierra, Decagon, etc. identify as generative AI companies. Their sole purpose is to deploy generative AI and limit themselves. You may have a bazooka, but if your enemy is near you, you need a knife. This is the reality of the contact center, where you need multiple tools,” he said.

Omilia itself has expanded its operations to building self-learning agents that can work across different customer contact points. To build that out, the company has now raised a $67 million Series B led by Expedition Growth Capital.

This round is the company’s second time raising capital after it raised $20 million from Grafton Capital in 2020. Since then, it has managed to increase its annual recurring revenue by 10x to $60 million.

Vassos says Omilia’s key differentiation lies in having great unit economics for both itself and its customers, and thanks to this approach, it hasn’t needed large cash infusions as compared to its competitors.

“In the next few years, we will see the companies that can offer real ROI prevail. We don’t mind that we’re not as sexy as ElevenLabs and Sierra on LinkedIn right now. We care about growing steadily and building the foundations for a billion-dollar revenue company in the next three years,” Vassos said.

Omilia’s clients include Capital One, Discover, RBC, DWP, and PSEG, and Vassos said quick-service restaurants that have adopted voice-based ordering are now a big focus. The company already has Taco Bell as a client, and has deployed its tech across more than 1,000 outlets. Vassos said that it is in talks with two more quick-service restaurants in the U.S.

But AI deployments can be quite hit or miss. In Taco Bell’s case, a snafu in its ordering system reportedly allowed a customer to order 18,000 cups of water last year.

Vassos maintains this incident never took place, and Omilia’s logs didn’t show it. We have reached out to Taco Bell to clarify, and we will update our story if we hear back.

Omilia will use the fresh cash to open a new office in the U.S., a market that accounts for a significant chunk of its revenue. It will also bolster its go-to-market team, and is hiring a chief revenue officer, chief marketing officer, and a VP of revenue operations.

The company has around 500 employees at the moment and expects its headcount to reach 600 by the end of the year.