TheChek News Flow

Here’s a news flash: The quasi-revenue numbers for OpenAI that got wide circulation last week were off— way off. Instead of $70 billion in annualized revenue, as we and others reported, OpenAI is generating closer to $50 billion in annualized revenue (sometimes shortened to ARR). A Financial Times report of the lower number on Thursday proved such a shock to investors that stocks of chip firms and neoclouds, including Nvidia, CoreWeave, Nebius and others, fell on the news, CNBC reported.

The cause of this debacle is complicated. But it’s a reminder that we all rely too much on annualized revenue as a way to track a company’s performance. It’s become a standard measure for AI companies because their revenue has been growing by leaps and bounds, making the more common quarterly reports seem a little quaint in their backward-looking nature. But taking one month’s revenue and multiplying it by 12, as is the standard way these annualized-revenue numbers are calculated, is a poor substitute for a fuller picture of a company’s performance. (This isn’t a new issue—see this 2020 story on the subject.)

We don’t know enough about what happens in that one month to know whether we can accurately extrapolate it by 12. The risk is compounded by differences between OpenAI and Anthropic in how revenue is calculated, particularly whether it includes the cut paid to cloud firms. (See our report on today’s developments).

Then there’s the fact that companies publish these numbers in very occasional blog posts, but not on any kind of regular basis. Much of the time news reports about ARR numbers appear based on communications with investors, which then find their way to reporters. And, as occurred with the mistaken $70 billion figure, sometimes investors and news outlets do clunky calculations that lead to errors.

This issue should recede when the AI firms go public. That doesn’t mean they’ll stop issuing annualized-revenue estimates, as some public companies have started to do. But at least we’ll see the real numbers and be better equipped to assess what’s going on. Right now—aside from the occasional leak of detailed financial statements—we’re in the dark.

Semiconductor M&A has been humming. One of the next deals could be for Eliyan, whose technology allows separate chips inside an AI system to talk to one another more quickly and efficiently.

The five-year-old startup, valued at $1 billion in July, has notified some of its investors in the last few days that it received a takeover bid, according to people familiar with the matter. It has been working with an investment bank since the summer and is likely to aim for a valuation of $3 billion, according to some of the people.

Softbank-owned Arm, an existing investor, held early talks about potentially buying Eliyan, but it didn’t make a bid and would be unlikely to do so at that valuation, said a person familiar with the matter.

The takeover interest follows big exits for other chip firms, such as Nvidia’s $20 billion licensing deal for Groq and Cerebras’ $56 billion IPO, as well as high-valuation funding rounds, such as Etched’s $21 billion fundraise. This has lifted founders’ hopes of a rich exit.

Spokespeople for Eliyan and Arm declined to comment.

Founded in 2021, Eliyan has raised more than $250 million from some of the world’s largest semiconductor and technology companies, such as AMD, Cisco and Meta, as well as venture firms Seligman Ventures and Tiger Global Management.

Eliyan is developing technology that helps large chipmakers solve a key bottleneck in AI systems: the traditional copper links that chiplets use to send data to each other and to high-bandwidth memory can be slow. That can leave AI chips idle no matter how capable and fast they are. Eliyan says its product solves some of the problems with copper with a solution that uses light and electronics.

“People are looking for a replacement for copper that has limited distance, and we’ve come up with an optical solution that provides all the features of a copper interconnect in terms of reliability, power, cost, but much longer reach... to connect these AI racks together,” Eliyan co-founder Ramin Farjadrad told The Information’s TITV in July.

Large buyers have been snapping up chip components makers in a bid to make AI chips run more efficiently. In February, Marvell Technologies bought Celestial AI, which develops optical interconnect technology, for up to $5.5 billion, even though the smaller company hadn’t generated any revenue at the time of the deal. In July, Analog Devices bought Empower Semiconductor, whose devices help AI chips manage electricity more efficiently, for $1.5 billion.

Amazon in August also raised the prospect of a similar integration involving Nvidia GPUs and Amazon Trainium AI chips, saying it would be able to use both chips in the same data center with similarly-structured server racks. It also said Trainium chips would tap Nvidia’s networking and memory hardware.

Another chip startup, SambaNova Systems, also demonstrated a way to connect its chips with Nvidia GPUs to jointly run AI models in June. Those integrations are happening as Nvidia separately develops a chip system that combines GPUs with specialized AI inference chips from Groq, whose founders and other employees Nvidia hired in a $20 billion technology licensing deal.

The Department of Justice is investigating whether Nvidia structured the Groq deal that way to avoid a formal review by the agency of the deal’s impact on the chip market, said people familiar with the probe, which was first reported by the New York Times.

Using chips made by different designers to handle the same AI workloads is a fairly new concept, sometimes referred to as disaggregated inference. Even as it pursues disaggregated inference with smaller rivals, Nvidia publicly says its own chips, on their own, are still the most efficient for handling nearly all AI workloads.

Still, cloud providers and major AI developers such as Anthropic and OpenAI don’t want to rely on just one chip vendor. And those two AI firms are separately developing their own chips for AI inference as well as using chips from other firms to lessen their reliance on Nvidia’s.

D-Matrix has focused on powering a process known as speculative decoding that speeds up an AI model’s performance. The forthcoming system with Nvidia will use d-Matrix chips to run a small AI model that guesses what a bigger model’s answer to the customer should be, and the bigger model will run on an Nvidia GPU to verify and accept the guesses.

The Information previously reported that d-Matrix was targeting new funding at a $5 billion valuation, though it’s unclear whether the capital from Nvidia would be part of that round. D-Matrix last raised $275 million at a $2 billion valuation in November 2025. D-Matrix started producing its first chips over the summer and is now testing its second generation chips.

Zach Yadegari, the teen co-founder of calorie tracking app Cal AI, has launched Persona, a new AI agent startup, and already raised $10 million. Vine Ventures led the round, with Z Fellows founder Cory Levy and Collective Global participating.

In March, MyFitnessPal acquired Cal AI for an undisclosed sum after it zoomed ahead of the older calorie app in the app store rankings. The deal was big enough to convince Yadegari and his co-founders to sell the company after it had pulled in over $30 million in annual revenue in under two years, MyFitnessPal told TechCrunch at the time.

Yadegari and his co-founders originally went to work for their new owner, continuing development on Cal AI, but he left the bigger company in June, he tells TechCrunch. He then went right to work on his next idea. (He is working with another co-founder who has yet to be announced.)

That idea is Persona, a personal AI assistant similar to Instinct or Meta’s Muse but with a hardware component that will soon be available: a $179 wearable band.

That makes Persona somewhat more like Bee, the wearable AI device that Amazon acquired in 2025, but Persona’s device differs in several ways, Yadegari tells TechCrunch. For one, it won’t be ambient AI, meaning it won’t always be listening and recording. The band will have a button that activates the assistant, or it can be set up to activate with the flick of one’s wrist.

Yadegari also says that Persona is designed with user privacy in mind. Persona doesn’t run its AI model or store user data on the device. Instead, the model runs in Persona’s cloud, and user data is sent there. The assistant protects privacy by encrypting data in transit and at rest. Users can also delete their data, the company says.

In this way, it is promising that user conversations stay private and cannot be sold to advertisers or data brokers.

The bands are expected to be available in December, and an app is in the works. In the meantime, Persona has launched as a free beta product in which users work with the assistant via iMessage text messages. It has attracted “a few thousand beta users,” Yadegari said. The band has also already generated “five figures in preorder revenue.”

Like other consumer AI assistants, he says, Persona can order food or rides, take notes, book travel, answer questions and emails, among other tasks.

“All of the utility apps that you use on your phone day-to-day, whether those are apps for shopping, whether it’s Uber, DoorDash, your email — these are all collapsing and will collapse into a single super app powered by an AI assistant,” Yadegari said.

Eventually, he hopes to make Persona even more proactive. “It can offer help and solve problems before you even know you have a problem, and overall be more seamless with your life, so that you don’t need to stare at a rectangle in your hand all day.”

Come see Zach Yadegari speak at TechCrunch Disrupt on “How to create viral growth and capitalize on it.” Disrupt takes place October 13-15 at the Moscone Center in San Francisco. Before Cal AI, Yadegari built his first business in the ninth grade and sold it for $100,000 when he was 16.

Investors continue putting money into Arini’s $7.3 billion flagship fund, betting rising rates will create more opportunities for founder Hamza Lemssouguer

Arini Capital, the hedge fund launched by a star trader known for concentrated bets, lost about 15.7% this year through September in its flagship fund after prices dropped on several of its largest investments, according to people familiar with the matter.

The London-based firm specializes in corporate debt and has been one of Europe’s fastest-growing fund managers, amassing around $22 billion since founder Hamza Lemssouguer, a former corporate-debt trader at Credit Suisse, launched it in 2022.

The 36-year-old fund manager is now facing the possibility of his flagship fund’s first annual loss unless one or more of its biggest investments turn around in the next few months.

Arini’s $7.3 billion flagship fund lost 7.44% in September as prices dropped on bonds and loans it owns in distressed companies such as telecommunications providers Altice International and Brightspeed, as well as carmaker Aston Martin AML -0.66%decrease; down pointing triangle, the people familiar with the matter said.

Altice bond prices dropped about 28% since July as creditors fought owner Patrick Drahi for control while Aston Martin’s debt fell a similar amount amid contentious debt restructuring talks.

Known for his math skills and obsession with exotic parrots, Lemssouguer has faced big swings before, given his penchant for taking large stakes in companies in turmoil. The flagship fund lost 16% during its first year before rebounding to a 4% gain and lost about 8% in early 2024 before finishing the year up 20.8%.

In a recent letter to fund investors, Arini said such volatility was to be expected.

“We do not believe it is an indicator of soured investments, but rather part of the process—being contrarian is a necessary condition for this opportunity set,” Arini wrote. “The periods of greatest opportunity can also involve substantial mark-to-market volatility.”

The Financial Times earlier reported on the performance of Arini’s flagship fund.

Despite the downturn, outside investors continue to buy into Arini’s flagship fund. The fund took in $425 million in October, the people familiar with the matter said.

Rising interest rates have intensified pressure on the heavily indebted companies, pushing prices of their debt down and creating bargains for distressed investors. Arini is one of the few such specialists in Europe, where high fuel prices and rising government bond yields in countries like France are straining companies even further.

Arini’s funds that lend to more stable companies have delivered better returns than the main fund this year, reflecting strong market demand for debt with credit ratings above the bottom bracket. The asset manager’s credit opportunities fund has returned 9% through September.

Elon Musk’s space company is buying spectrum licenses to help build a mobile network to compete with telecom giants

SpaceX SPCX -4.19%decrease; down pointing triangle reached a deal to buy cellular spectrum licenses from investment firm Grain Management, its biggest move yet into a wireless market dominated by traditional telecom companies.

The Elon Musk-led space company will pay about $8 billion in cash for the assets, according to people familiar with the matter. Grain and SpaceX SPCX -4.19%decrease; down pointing triangle had no comment on the value of the transaction.

The U.S. licenses, which are in the 800-megahertz band, are tailored for wireless service from cellphone towers rather than satellites. The spectrum could serve SpaceX’s stated ambition of moving its Starlink satellite service beyond broadband into the wireless market to compete with the likes of AT&T T 1.63%increase; up pointing triangle and Verizon.

U.S. cellphone carrier stocks sank on the news. Shares in AT&T, Verizon and T-Mobile all fell more than 6% in post-close trading. SpaceX shares climbed about 3% to $165.10 after hours.

The Federal Communications Commission earlier this week approved SpaceX’s bid to launch up to 15,000 satellites dedicated to mobile service.

The company said that upon FCC approval of the spectrum purchase, it will combine the growing constellation with “an advanced terrestrial deployment” to connect devices. It said its new spectrum will help cover places satellite links can’t reach, including inside buildings.

SpaceX has offered basic wireless connections in remote areas through a T-Mobile partnership for more than a year. But it has been upgrading its Starlink Mobile network with newer satellites capable of handling more data.

Starlink still needs some terrestrial infrastructure, such as cell towers or rooftop equipment, to use the spectrum for mobile coverage, said Roger Entner, founder of telecom research firm Recon Analytics, whose clients include AT&T, Verizon and T-Mobile US.

Two months ago, SpaceX executives floated the idea of using small cells rather than traditional cell towers to support a mobile network. Building a system with traditional towers would likely cost billions of dollars to provide basic coverage, analysts have said.

SpaceX made an early commitment to the wireless market last year, reaching a roughly $17 billion cash-and-stock deal to obtain satellite-based spectrum rights from EchoStar. The purchase gave Musk’s company a platform for beaming data to cellphones from space using its own reserved frequencies rather than through airwaves licensed from partner T-Mobile.

EchoStar later agreed to sell some ground-based cellular licenses to SpaceX for $2.6 billion. The purchase hinted at SpaceX’s budding interest in wireless service, though the spectrum alone wasn’t enough to launch a full-fledged cellphone network.

AT&T, Verizon and T-Mobile earlier this year teamed up to extend their service to remote areas over satellite links, a move widely viewed as a defensive tactic against potential SpaceX encroachment.

Machines programmed to sort merchandise in warehouses hit kinks with cardboard boxes and turkeys; automation push reaches ‘peak complexity’

Walmart WMT 2.22%increase; up pointing triangle has relied for years on more than 140,000 warehouse workers to unload trucks, forklift pallets of merchandise into warehouses, sort and store the goods on towering racks, then reverse the entire flow to send them out to stores and homes.

It isn’t an easy job. A worker might walk 20,000 steps a shift. It also isn’t all that complicated—unless you try to program robots to do it.

Consider the seemingly simple task called “breakpack”—removing items from cardboard boxes to send out individually. It’s a no-brainer for Walmart’s humans. But it turned into one of the many challenges bedeviling the company’s decadelong, multibillion-dollar quest to automate its warehouses.

Walmart and the robotics companies it hired to help searched in vain for a way to choreograph robots to sort merchandise into bins while zipping past warehouse workers, testing technology that either didn’t work or cost too much. Eventually, they designed small robots on wheels with motion and obstacle sensors and are now installing them in more than a dozen warehouses.

“We are kind of in this peak complexity,” said Rob Montgomery, head of supply-chain operations for Walmart U.S., who is overseeing the automation transition now under way at most of Walmart’s roughly 200 U.S. supply-chain buildings.

Problems have multiplied. The cardboard shipping boxes Walmart has used for years are too big for the automated systems. The miniature robots that shuttle goods break down too often. The systems can’t accommodate some oversize products, such as big bags of dog food.

What makes the automation campaign so tricky, Montgomery said, is that Walmart is re-engineering its warehouses without taking them out of service.

Many of the nation’s largest retailers are struggling through similar automation transitions as pressure grows to boost sales and ship merchandise to homes faster without spending a fortune on additional workers or facilities. The rapid growth of Amazon—which is ahead of its retail rivals on automation and earlier this year surpassed Walmart to become the largest U.S. company by annual revenue—has dialed up the urgency.

“In many cases, investments in automation are an arms race” between retailers, said Rueben Scriven, senior warehouse automation research manager at market-intelligence firm Interact Analysis. “And then the Amazon effect is the big elephant in the room.”

Retailers without the capital to invest heavily through a long period of troubleshooting are at a disadvantage. Last fall, Kroger said it would take a $2.6 billion charge to close three automated e-commerce fulfillment centers as a result of “the automated fulfillment network not meeting financial expectations.” Kroger said it would use stores more often to fulfill online orders and focus on “capital-light” automation efforts.

Walmart expects spending on remodeling stores and rolling out supply-chain automation to peak this year and next. Company executives have said the automation push means it eventually will need fewer warehouse workers, and more of the jobs that remain will involve managing or fixing automated systems rather than lifting boxes.

Because turnover is high in warehouse jobs, the executives said, automation won’t necessitate layoffs. “We still need people that unload trailers,” said Montgomery, adding that overall sales growth could require as many warehouse workers as today.

Walmart executives said the automation should allow the company to process more inventory, faster, and with more precision, reducing the frequency of out-of-stock goods and reducing costs. They envision a time when a choreographed and predictable cadence of trucks arrive at stores loaded with pallets of goods needed at that moment. The pallets will be organized so store workers can pull them to a single aisle, and the items can be unloaded in the order they are stocked on shelves. That would make store worker schedules more precise and predictable.

First, the company needs to work out lots of kinks.

Technical challenges

For more than two decades, Walmart has used the same kind of cardboard box to shuttle many goods between warehouses and stores, stamped with a reminder that they cost the company an average of $1 and should be reused. Walmart plans to throw them all out in coming months. The new ones are 1½ inches shorter and need to be taped closed, an added step that will cost money.

“It’s a small change,” said Phillip McIlrath, the manager of an Indiana apparel warehouse using the new boxes. “But associates doing the work down here, it’s changing their daily routine quite a bit.”

The new boxes, said Montgomery, will reduce stocking time at stores, and hence labor costs.

Walmart’s supply chain feeds 5,200 U.S. stores that sell everything from groceries to clothing to tires, both off the shelves and online. The system, Montgomery said, needs to “handle almost anything that we could put into it, from any supplier and at any time. That has been a challenge.”

Many of the nonperishable goods on store shelves are first delivered to warehouses that consolidate shipments from suppliers, then sent to 42 distribution centers. Dozens of separate grocery warehouses handle cold and frozen items. Still others handle many of Walmart’s U.S. e-commerce orders, as do the backrooms of stores.

Most of Walmart’s supply-chain buildings were built in the 1980s and ’90s, so about 80% of its automation construction projects are retrofits. The company also has built several new automated grocery and e-commerce warehouses from scratch.

Different automated systems are needed to handle different items at different locations, often built by different companies. One system is designed to handle full cases of goods sent to Walmart by suppliers, another to sort individual items from their shipping cases. Other systems handle fresh and frozen food or e-commerce orders.

Walmart started trying versions of the systems more than a decade ago, often in a 1.6 million-square-foot warehouse in Brooksville, Fla., that processes nonperishable goods. Those systems are made by Symbotic, a Wilmington, Mass.-based robotics company in which Walmart has a 12.6% stake.

Walmart and Symbotic struggled to come up with a system that worked well enough to install in other locations. In 2019, they ripped one version out of Brooksville and rebuilt it from scratch. In 2022, Symbotic announced that it had struck a deal with Walmart to expand its systems to all 42 warehouses for nonperishable goods by around 2030.

The system stores goods in floor-to-ceiling, open-air metal structures where robots shuttle goods between truck unloading areas and storage racks, and other robots with arms that build 10-foot-tall pallets that are loaded onto trucks bound for stores.

For Symbotic, it was difficult to hit Walmart’s goals, primarily the desired “throughput,” the rate at which goods are processed, said people familiar with the situation. Symbotic tracks the performance of its equipment on an electronic board inside its Massachusetts headquarters, hoping to snuff out problems quickly.

Automating Walmart’s warehouses “takes time, hard work, and a strong partnership defined by trust,” said Symbotic CEO Rick Cohen in an emailed statement.

Robot issues

A persistent focus has been the frequency with which Symbotic’s product-shuffling robots break down or need maintenance. The robots can be thwarted by something as simple as dust buildup, and stoppages can create a cascade of other problems, said some of the people familiar with the situation.

Last year, under 10% of the robots were being fixed at any given time, said a Symbotic spokesman. The percentage has dropped to about 5% in recent months, he said.

From the start, one of Walmart’s primary automation goals has been building a neat, tightly wrapped pallet in the exact order that it needs to be unloaded at stores, which would save on labor costs. Walmart and Symbotic haven’t achieved the goal yet, but have made progress, said executives at both companies.

As more products move through automated warehouses, the Tetris-like task of building pallets should become more precise because software will have more items to choose from, said Montgomery.

Automation is driving up some costs. The monthly electric bill at Walmart’s automated grocery facilities can be $800,000 or more during peak cooling or warming months, up from about $250,000 a month when operations were more manual, said a person familiar with the situation. The mechanical and electrical engineers needed to run the operations cost more than unskilled warehouse workers.

Earlier this year, Walmart was behind on the automation targets it laid out in 2023, when investors and journalists toured its Brooksville automation test warehouse. At the time, executives said by the end of January of this year, about 65% of stores would be serviced by warehouses using automation and 55% of e-commerce warehouse volume would be automated. By August, Walmart was close: More than 65% of stores were receiving some automated freight and over 50% of e-commerce inventory was moving through automated systems.

Brooksville is now nearly “fully automated,” said Montgomery, which means about 80% of inventory moves that way. The other 20% is trickier. Items such as large-screen TVs and jumbo bags of dog food will need manual assistance for the foreseeable future, he said. Overall, he said, the company now has a handle on the technology it needs to automate.

Some Walmart suppliers need to adapt their products to work in the automated systems. A few years ago, during the run-up to the Thanksgiving holiday rush, workers in one of the new automated California grocery warehouses realized that its robotic arms fumbled large frozen turkeys. Walmart asked its supplier to reinforce their shipping boxes. Amid a last-minute spike in demand, nearby grocery warehouses were used to ship enough turkeys to stores.

Earlier this year, Walmart started quickly informing suppliers when and how their packaging presents problems.

Walmart hasn’t figured out the best way to automate online orders assembled inside stores, not warehouses. It has invested nearly $1 billion since 2016 trying.

Automating e-commerce in stores is hard because space is tighter. And snafus can be maddening for customers who might be waiting inside cars in the parking lot or expecting home delivery at a specific time.

Walmart introduced a system built by Alert Innovation, another Boston-based robotics company, to 20 stores. But Alert struggled to meet Walmart’s benchmarks on throughput and other measures and faced a cash crunch, according to people familiar with the matter. In 2023, Walmart bought Alert for $400 million.

Walmart then sold the company to Symbotic last year at a loss for $200 million. Symbotic is currently installing a new test version in a Dallas Walmart and running the existing 19 locations, but doesn’t expect to move forward with a workable model until 2028, Symbotic executives said on a recent conference call to discuss earnings.

Montgomery said Walmart’s leadership team expects there will be messy moments, but top executives and the board view automation as a long-term project that’s worth it.

“We know that if we wait till perfect,” he said, “it’ll take too long.”

A Starbucks-Chipotle merger offers the wrong kind of synergy

The overlap between the two brands is unlikely to be enough to justify a takeover premium

Putting the band back together is surely a great feeling, but a dubious corporate strategy. Starbucks chief Brian Niccol has considered a merger with Chipotle, his $41bn rival — and former employer — the FT reported on Thursday. However neat the match might seem from Starbucks’ corner office, it makes little sense for shareholders.

Niccol knows Chipotle very well indeed. He ran the Mexican-themed restaurant chain for six years before decamping in 2024 to the world’s biggest coffee chain, which was then in a state of considerable disarray. Seen from afar, the two do faintly similar things, serving comestibles to customers who are in a hurry. Both have large, sometimes restive workforces.

Moreover, the ways in which they differ might seem from a distance like opportunities. Chipotle’s balance sheet is free of debt; gearing it up a little could provide low-hanging fruit. And while Starbucks has a sprawling global network, Chipotle has yet to venture outside the US, apart from a smattering of international stores.

But do these companies really belong together? Not really. For one, the overlap is unlikely to be enough to justify a takeover premium. Imagine Niccol could shave 10 per cent off Chipotle’s food, drink and packaging costs and halve its general overheads. That would save $770mn a year, Lex reckons using Royal Bank of Canada forecasts, giving an after-tax return on investment of just 6 per cent. Anything below 10 per cent should make investors nervous.

More importantly, Niccol hasn’t finished fixing up his present employer, making this a poor time to take on his old one. Starbucks shares have gone sideways since the day his appointment was announced. Profitability isn’t yet where it used to be. Same-store sales growth of 8 per cent in the latest quarter suggests the company has turned a corner, but recent numbers are still flattered by the previous year’s decline.

Chipotle has fared much worse, as has most of the US restaurant sector. But for Niccol to argue he can do a better job than its current leadership might be somewhat problematic, since Chipotle chief Scott Boatwright was his direct report for six years. Conversely, keeping the brand separate and its leaders in place — the Starbucks boss spent more than a decade at multi-brand restaurant group Yum Brands — would further trim the potential cost savings.

Perhaps it’s just that Niccol could use some more motivation. After all, the generous $100mn-plus package he received in 2024, linked to the company’s share price, has mostly failed to pay out. Worse, having started his role in sunny California, he now spends significant time in rainy Seattle. Starbucks should want to keep Niccol engaged, but buying Chipotle is too high a price.

F1 is shifting away from Europe. Can Britain keep what’s left?

Fans mourn the loss of historic races, but they are looking at the wrong map. The one that matters shows a strip of English countryside that Britain takes dangerously for granted

Over the past decade, Formula 1’s race map has been quietly redrawn, with its European heartland shrinking as owner Liberty Media globalises the sport and pursues higher revenues.

Europe has gained Madrid in September — with the Spanish capital returning to the F1 calendar after 45 years — and, next season, Portugal’s Portimão. But it has lost Italy’s Imola and will lose Zandvoort in the Netherlands from 2027; while Belgium’s Spa-Francorchamps, one of the sport’s iconic circuits, will alternate annually with Barcelona in Spain.

The return of the Turkish Grand Prix in 2027 — held on the Asian side of Istanbul — means Asia will host nine races next season, overtaking Europe with eight, for the first time.

While F1 CEO Stefano Domenicali has reassured fans that the German Grand Prix will return, many traditionalists see a clear trend: Europe is losing its primacy as F1 embraces higher-bidding hosts.

But they are grieving over the wrong map.

F1 has always had two geographies. The first is where the show takes place: the races. That geography is globalising fast, and understandably so: Gulf hosts pay Liberty Media roughly $50mn a year for their races, according to analyst estimates, double that of most European promoters.

The second is where the cars are made: the industry. That geography is historically clustered. Since the second world war the design and manufacturing of F1 cars has concentrated in a strip of English countryside between Oxford and Northampton.

Although Italy’s Motor Valley, in Emilia-Romagna, remains home to Ferrari, Racing Bulls and Dallara in F1, the British cluster remains the sport’s main centre of gravity for now.

Motorsport Valley, as the strip is known, is home to the headquarters of six of the 11 teams: Mercedes in Brackley, Red Bull in Milton Keynes, Aston Martin at Silverstone, Williams in Grove, Alpine in Enstone and McLaren in Woking. Even teams headquartered outside the UK keep a foot in the Valley: Haas runs key operations from Banbury; Racing Bulls has a technical centre in Milton Keynes; while newcomers Cadillac and Audi have established a UK base at Silverstone and a technology centre in Bicester, respectively. The Motorsport Industry Association values the wider UK motorsport and engineering services sector at £16bn a year and estimates that it supports 50,000 jobs.

None of this was planned. The cluster’s origins can be traced back to amateur enthusiasts racing home-built cars on disused wartime airfields, Silverstone among them. Charles Cooper and his son John built tiny rear-engined racers in a garage in Surbiton, winning back-to-back F1 world championships in 1959 and 1960. Colin Chapman assembled his first Lotus in a north London lock-up and went on to claim seven constructors’ titles.

Two mechanisms still power the Valley today. Talent rises vertically, from karting paddocks and club racing into professional teams; and it moves horizontally, as engineers circulate between rivals, spreading knowledge that no patent can capture. Passion anchors people to place. That is what makes clusters sticky, and why this one has survived every change of regulation, ownership and fashion since the 1950s.

It is also why the anguish over the calendar misses the point. Hosting races and building cars are different markets: one sells spectacle to a global audience; the other drives economic value through advanced engineering with benefits and knowledge transfer beyond F1 and motorsport.

Italy proves the point. It lost Imola, yet its production base thrives: Ferrari in Maranello, Racing Bulls in Faenza and Dallara in Varano de’ Melegari, which designs Haas’s F1 chassis, builds Cadillac’s endurance prototypes and supplies every F2, F3 and IndyCar.

A country can lose the show and keep the factory. Italy still has both, and England more so: most of the teams and suppliers, F1’s London headquarters, and the British Grand Prix at Silverstone until 2034. The question is whether its engineering cluster can remain world-leading.

History can change, leaving no room or time for complacency. Motorsport Valley’s grip is already loosening: from 2029, General Motors will build Cadillac’s engines in North Carolina; Audi has based its works programme in continental Europe; and Dallara keeps winning outsourced work that teams choose not to do in-house — Haas and Cadillac being two points in case.

Meanwhile, UK frictions are growing, with higher living costs than elsewhere deterring international talent from moving to the country, along with rising costs of energy, manufacturing and the visas for the foreign engineers that are a key part of any F1 team. Clusters rarely collapse loudly. They erode quietly, one outsourced contract at a time. Ask Detroit.

The solution for British policymakers is to treat Motorsport Valley as strategic infrastructure rather than a heritage curiosity. The government already offers fast-track visas for digital technology and the arts, and it waves through elite drivers as international sportspeople. However, particularly after Brexit, the engineers who make those drivers and their teams champions must queue with everyone else. Extending endorsement to motorsport engineering would be a powerful move.

Moreover, F1 research and development should enjoy similar levels of state support as the aerospace industry, and the government should support the amateur layer of the pyramid, such as karting. My research suggests governments cannot legislate a cluster into existence; passion does that. But policy determines whether the next Cooper garage starts in Surbiton, Concord or Varano.