On Tuesday morning, 7 April 2026, Intel announced it was joining Elon Musk's Terafab project — and in doing so, quietly revealed the truth that Musk's theatrical March launch had obscured: Tesla is the client. Intel is the foundry. The rest is showmanship. "Intel is proud to join the Terafab project with @SpaceX, @xAI, and @Tesla to help refactor silicon fab technology," the company posted on X, adding cheerfully that it had been "fun hosting @elonmusk at Intel this past weekend." A weekend factory tour and a single social media post, and suddenly the most audacious chip project in the history of semiconductor manufacturing has a co-pilot with 58 years of fabrication experience, a 40,000-wafer-per-month facility already running in Arizona, and a $37.4 billion cash position.
The word doing the heaviest lifting in Intel's announcement is not "Terafab." It is not "SpaceX." It is not even "1 terawatt" — that intoxicating, borderline-fictional unit of compute ambition. It is Intel's description of its own contribution: the ability to "design, fabricate, and package ultra-high-performance chips at scale." That is not the language of a supplier. That is not the language of a licensing partner. That is the language of a company announcing that it runs the factory.
The Most Epic Chip Exercise in History — With a Footnote
To understand why this matters, rewind seventeen days. On the night of 21 March 2026, Musk walked onto a stage inside the defunct Seaholm Power Plant in downtown Austin — because of course he did, because nothing says "industrial revolution" quite like an abandoned power station — and announced Terafab to a crowd that included Texas Governor Greg Abbott and a sky full of light beams. He called it "the most epic chip-building exercise in history by far." He said his companies needed chips at a rate the global semiconductor industry simply could not supply. He said, with the calm certainty of a man who has landed rockets on drone ships in the Atlantic, "We either build the Terafab or we don't have the chips, and we need the chips, so we build the Terafab."
The pitch was intoxicating. Terafab, as Musk described it, would be a vertically integrated mega-facility on the North Campus of Giga Texas in Austin. Every stage of semiconductor production — chip design, lithography, fabrication, memory production, advanced packaging, and testing — under one roof, in one building, for the first time in the history of chipmaking. The target: one terawatt of artificial intelligence compute capacity per year. The process node: 2-nanometer class, the sharpest edge of what human beings can currently manufacture. The initial output: 100,000 wafer starts per month, scaling eventually to 1 million. The cost: Rs 1.7 to 2.1 trillion ($20 to $25 billion) for the pilot phase, not yet incorporated into Tesla's 2026 capital expenditure plan. The long-term vision: 100 to 200 billion custom AI and memory chips per year. The stated motivation: Tesla, SpaceX, and xAI together consume AI compute at a rate the rest of the world's foundry capacity delivers at roughly 2 per cent.
"There's a maximum rate at which they're comfortable expanding," Musk said of TSMC and Samsung. "That rate is much less than we would like." The audience cheered. The internet did its thing. TSMC's share price barely flinched, because TSMC knows something the crowd in Austin did not: building a leading-edge semiconductor fab from scratch takes four to five years under ideal conditions, requires lithography machines that ASML produces at the rate of roughly 179 per year globally, and demands a workforce of specialists that would outnumber the entire graduating class of every electrical engineering programme in America for a decade.
But Musk's supply problem is entirely real, and the urgency behind Terafab is not theatre. Tesla's AI5 chip — the fifth-generation processor that will power Full Self-Driving, the Cybercab robotaxi, and the Optimus humanoid robot — is currently manufactured by TSMC in Arizona and Taiwan, with small-batch production expected in late 2026 and volume production targeted for mid-2027. The AI6, designed for enterprise-scale AI workloads and promising roughly twice the AI5's performance, is under a Rs 1.38 trillion ($16.5 billion), eight-year supply agreement with Samsung's new 2nm fabrication facility in Taylor, Texas. And the AI7 — which Musk has described as "more adventurous" — will need different fabs entirely. Terafab is that different fab. Intel is what makes it possible.
The Crisis Nobody Is Talking About
Before examining what Intel brings to Terafab, consider the supply-chain fracture that made this partnership not merely attractive but structurally necessary. Samsung's 2-nanometer Gate-All-Around process — the technology underpinning the AI6 contract — has slipped by approximately six months. A postponed multi-project wafer run has pushed the AI6's mass production timeline from mid-2027 to late 2027, compounding a pattern of delays that has already pushed the AI5 volume production date back from late 2025 to mid-2027 after Musk declared the design "finished" in July 2025 and then "almost done" again in January 2026.
The Samsung delay is not a Tesla-specific problem. It affects every customer on Samsung's 2nm process, including South Korean AI chip startup DeepX. Samsung's foundry division has been loss-making and under-utilised, and the Taylor, Texas facility — one of two Samsung is building in Texas as part of a Rs 3.36 trillion ($40 billion) state investment — postponed its completion once already, originally to upgrade from 4nm to 2nm. "Their foundry business has been loss-making and struggling with under-utilisation, so this will help a lot," said Vey-Sern Ling, managing director at Union Bancaire Privee in Singapore, when the AI6 deal was announced in July 2025. That assessment looks rather more complicated now that the 2nm process is running six months behind schedule.
According to Counterpoint Research's Foundry Market Supply Tracker, the global semiconductor foundry market generated a record $320 billion (Rs 26.88 trillion) in revenue in 2025, growing 16 per cent year-over-year. TSMC accounted for 38 per cent of that total and grew 36 per cent — four times the rate of its nearest competitors. Samsung had, as Counterpoint described it, "a mixed year." Intel Foundry held 6 per cent of the broader Foundry 2.0 market by revenue. IDC's Galen Zeng, senior research manager for semiconductor research at IDC Asia/Pacific, placed the macro context in sharp relief in a report published 2 April 2026: "The Foundry 2.0 market is entering a steady expansion cycle driven by AI in 2026. Advanced nodes and advanced packaging remain in short supply." IDC projects the entire foundry market to surpass $360 billion in 2026, growing 17 per cent year-over-year. The gap between first and second place in this industry is not measured in percentages. It is measured in process generations, and TSMC is two to three years ahead of everyone else.
This is the context in which Musk walked into Intel's facility last weekend. Not as a customer shopping for wafers. As a man with a burning house and a neighbour who keeps a fire extinguisher.
What Intel Actually Brings
Intel's post on X was precise in a way that corporate communications rarely manage to be. The company said its "ability to design, fabricate, and package ultra-high-performance chips at scale" would help accelerate Terafab's goals. That is three distinct capabilities in one sentence: design, fabricate, and package. This is the full semiconductor stack. This is everything a working fab requires. Tesla has chip designers. Tesla has ambitions. Tesla has money, though exactly whose money and how it is structured within Terafab remains undisclosed. What Tesla does not have is process engineers who have spent decades optimising lithography recipes, managing cleanroom yield variance, and building the institutional muscle memory that turns a billion-dollar machine into a functional factory.
Before moving past Intel's post, it is worth lingering on the single most carefully chosen word in it: "refactor." In software engineering, to refactor code means to rewrite its internal structure to make it cleaner and faster, without changing what it actually does. Intel did not say it would "build" a fab. It did not say it would "operate" one. It said it would help "refactor silicon fab technology." That could be a deliberate hedge — a way of claiming involvement without claiming ownership. But it could equally be something more interesting: a declaration that Intel views semiconductor manufacturing in the AI era as a software-like discipline, one that can be iterated, versioned, and improved at the speed of code rather than the speed of concrete. Musk's entire pitch for Terafab rests on that idea — compress the chip iteration cycle from nine months to days by putting design, fabrication, and testing under one roof. Intel, by using the word "refactor," may be signalling that it shares that philosophy. Or it may simply have a good social media team. Either way, the ambiguity is the story.
Intel has all of that, and it has it domestically, on American soil, at a moment when "American soil" carries the strategic weight of a geopolitical trump card. Intel's 18A process node — 1.8nm-class, the technology most likely to underpin Terafab's initial production — entered high-volume manufacturing at Fab 52 in Chandler, Arizona, in October 2025. The facility processes approximately 40,000 wafer starts per month at full ramp. Think of a wafer start as the moment a disc of pure silicon the size of a dinner plate enters the front of an extraordinarily complex factory; what emerges at the other end, weeks later, is a slice of that plate containing hundreds of chips, each one holding billions of transistors smaller than a coronavirus. Intel's 18A process packs those transistors using two proprietary breakthroughs: RibbonFET, which wraps the electrical gate around the transistor on all four sides for better control — like squeezing a garden hose from every direction instead of one — and PowerVia, which delivers power through the back of the chip wafer rather than the front, freeing up the surface for faster signal routing. It is, in the language of chip engineers, a generational leap. Yields remain below profitable levels and, per CFO David Zinsner, will not reach industry-standard thresholds until the end of 2026 at the earliest. Progress, but still progress in a tunnel.
Intel CEO Lip-Bu Tan, who took the helm in March 2025 after the ousting of Pat Gelsinger, set aside the caution in a separate X post on Tuesday: "Elon has a proven track record of reimagining entire industries. This is exactly what is needed in semiconductor manufacturing today. Terafab represents a step change in how silicon logic, memory and packaging will get built in the future. Intel is proud to be a partner and work closely with Elon on this highly strategic project." Tan had previously written in a memo to staff: "Going forward, our investment in Intel 14A will be based on confirmed customer commitments. There are no more blank checks. Every investment must make economic sense. We will build what our customers need, when they need it." A Tesla, SpaceX, and xAI anchor customer — collectively representing the most voracious AI compute demand of any private entity in history — is precisely the confirmed commitment Tan has been waiting for.
"Intel needs to show it can support the largest customers with their most important projects, and that seems to be the case with the Tesla partnership," said Gil Luria, analyst at D.A. Davidson, calling it an "important step" in Intel's restructuring. Intel Foundry's external revenue stood at $222 million in Q4 2025, a number that is, in the context of a $320 billion global foundry market, essentially symbolic. The symbolism of Terafab, however, is anything but.
Neil Shah, co-founder and vice president of research at Counterpoint Research, was among the first industry voices to weigh in publicly, posting on LinkedIn within hours of the announcement. He called it a "landmark win" for Intel and its foundry business, and mapped the implications on both sides of the deal. For Intel, he identified six strategic gains: credibility for its foundry capabilities, validation of what he termed the company's "IDM 2.5 Strategy" — running a huge-scale fab in a pure-play way — the addition of marquee customers for scaling Intel Foundry Services IP, co-innovation opportunities, expansion into space, robotics, and AI data centres, and a meaningful boost to investor confidence. For Tesla, SpaceX, and xAI, Shah pointed to supply-chain diversification beyond TSMC and Samsung, greater vertical integration through process co-innovation, and the geopolitical and policy tailwind of "Make in USA" alignment. He acknowledged that capital expenditure challenges "are galore" but said Intel's involvement addresses the most critical execution risk. Counterpoint Research, he noted, would follow with a deeper dive on foundry wafer capacity, yield, pricing, and customer models.
A Comeback Story Wearing A Rocket Ship's Clothes
To appreciate the full dramatic irony of Intel joining Terafab, it is worth briefly inhabiting Intel's recent past. The company that invented the microprocessor in 1971, that powered the personal computing revolution, that coined the phrase "Intel Inside" as a synonym for technological trust, spent the first half of this decade being publicly humiliated by a Taiwanese foundry and a Santa Clara GPU company. TSMC ate Intel's manufacturing lunch. Nvidia ate its data centre dinner. AMD ate the leftover sandwiches. By 2024, Intel had posted an $18.8 billion annual net loss — the kind of number that makes accountants stare at spreadsheets and quietly weep.
Then Lip-Bu Tan arrived. Tan, a venture capitalist turned semiconductor executive, flattened a management structure he found "eight or more layers deep," redrew reporting lines, and declared 2026 "an execution year." He raised $7 billion through private placements — $5 billion from Nvidia and $2 billion from SoftBank, a detail that sounds like a fever dream until you consider that Nvidia's business model depends on someone being able to manufacture chips at leading-edge nodes on American soil, which is increasingly a national security matter. The U.S. government converted its remaining CHIPS Act grants into a 9.9 per cent non-voting equity stake in Intel, making the federal government Intel's second-largest shareholder. Intel then bought back the 49 per cent stake in its Fab 34 in Ireland from Apollo Global Management for Rs 1.19 trillion ($14.2 billion), funded through cash on hand and $6.5 billion in new debt — a signal that the balance sheet, while stretched, now carries genuine conviction. Intel's stock is up 35 per cent year-to-date and 126 per cent over the past twelve months. KeyBanc analyst John Vinh had already raised his price target on Intel from $65 to $70, maintaining Overweight, the day before the Terafab news even landed — driven entirely by the Fab 34 buyback and 18A momentum.
Intel's Data Centre and AI Group delivered Q4 2025 revenue up 15 per cent sequentially — the fastest sequential growth in a decade, per Zinsner. Hyperscalers, Tan has said, are now telling Intel that "their first choice is the CPU from Intel." The proof that 18A works at server scale arrived in March 2026 in the form of Clearwater Forest — Intel's first data centre CPU built entirely on the new process. Think of it as Intel's first dish served from the new kitchen: a 288-core processor spread across twelve chiplets, all fabricated on 18A, stacked and connected using Intel's proprietary 3D packaging in a configuration whose engineering density would have been dismissed as science fiction from this company three years ago. Panther Lake, the first commercial 18A client chip, is now shipping in laptops. Intel's packaging technologies — Foveros Direct and EMIB — are genuinely world-class, a competitive advantage even TSMC struggles to match at the same integration density.
For Intel's foundry business, which has been searching for an anchor customer with the intensity of a lighthouse keeper scanning for ships, Terafab is the supertanker on the horizon.
Why the Timing Is Not a Coincidence
Six days before Intel's Terafab announcement, on 1 April 2026, SpaceX filed confidentially for an initial public offering with the Securities and Exchange Commission. The filing, confirmed independently by Bloomberg, CNBC, Reuters, and The Wall Street Journal within hours, puts SpaceX on track for a June Nasdaq listing at a target valuation of $1.75 trillion — a raise of up to Rs 6.3 trillion ($75 billion) that would make Saudi Aramco's 2019 IPO look like a Series A. Lead underwriters are Bank of America, Goldman Sachs, JPMorgan Chase, and Morgan Stanley. The public S-1 prospectus, which must drop at least fifteen days before the roadshow, is expected in late April or May.
SpaceX's $1.75 trillion valuation is anchored by Starlink, its satellite internet service, which ended 2025 with 9.2 million subscribers and over $10 billion in revenue, with projections reaching $22 billion by the end of 2026. The February 2026 all-stock merger with xAI, Musk's artificial intelligence venture — burning approximately $1 billion a month at the time of the deal — boosted the combined entity's pre-IPO valuation to $1.25 trillion. The Terafab Intel announcement, coming six days after the confidential filing, is not coincidental timing. It is narrative architecture. Before SpaceX sits down with institutional investors on a roadshow, it wants a story that says: we know how to build the semiconductor infrastructure for the AI and space age, and America's most experienced chip manufacturer is our partner. That is a considerably more compelling prospectus than "we make rockets and burn $1 billion a month on AI."
Musk himself has been explicit about the long-term vision that makes the IPO framing coherent. SpaceX has filed with the Federal Communications Commission for a licence to launch one million data centre satellites into low Earth orbit. At the Terafab launch event, Musk described a future where AI satellites are assembled on the Moon and launched into orbit using electromagnetic mass drivers — solar-powered railguns, essentially. He said: "I want us to live long enough to see the mass driver on the moon, because that's going to be incredibly epic." That is not a product roadmap. That is a civilisation roadmap. And Terafab — Intel-powered, Samsung-augmented, TSMC-supplied for near-term chips — is its first physical sentence.
3 Foundries, 1 Vision
The corrected picture of Tesla's silicon supply chain, post-Intel announcement, looks like this. The AI5, purpose-built for vehicle inference and described by Tesla as delivering 40 to 50 times more computing performance than the AI4 at approximately 10 per cent of the production cost of Nvidia's Blackwell chip: manufactured by TSMC, initially in Taiwan and transitioning to TSMC's Arizona facility. Small-batch production in late 2026; volume in mid-2027. The AI6, targeting enterprise-scale AI training and inference with roughly twice the AI5's performance: manufactured by Samsung's 2nm GAA process at Taylor, Texas, under the Rs 1.38 trillion ($16.5 billion) contract running through 2033 — delayed by approximately six months due to Samsung's production slip, now targeting late 2027. The AI7, which Musk has said "will need different fabs, as it is more adventurous": Terafab. Intel. Beyond 2028.
Terafab also produces the D3, a chip class that has received almost no coverage despite being, by Musk's own numbers, the facility's primary output. The D3 is a radiation-hardened processor designed for operation in low Earth orbit — built to withstand cosmic radiation, operate at higher temperatures than terrestrial chips to minimise radiator mass on satellites, and run AI inference workloads aboard SpaceX's planned orbital data centre constellation. Musk has stated that 80 per cent of Terafab's compute output will go to space. Twenty per cent for Earth. The ratio tells the real story of what Terafab is actually for: it is a space computing infrastructure project with a side business in cars and robots.
The AI5 will also come in "slightly different versions" at TSMC and Samsung, as Musk noted in a post on X, attributing this to differences in how each foundry translates chip designs into physical silicon. He said Tesla's AI software would behave identically on both versions. This dual-foundry approach for a single chip generation is, by semiconductor industry standards, unusual and expensive. It also tells you precisely how seriously Tesla takes supply-chain concentration risk after watching the AI4's production be entirely dependent on Samsung and knowing what a six-month production slip feels like when you have a robotaxi network to launch.
$5 Trillion Is Not a Typo
At this point, a reasonable person — or, for that matter, an unreasonable analyst at Bernstein — might raise their hand and note that Elon Musk has a somewhat complicated relationship with announced timelines. The Hyperloop is not operational. Full Self-Driving has been "one year away" for approximately eight years. The Cybertruck arrived, eventually, but not exactly on schedule. And Terafab at full scale — 1 terawatt of annual compute, 1 million wafer starts per month, 100 to 200 billion chips per year — is not a semiconductor facility. It is a semiconductor civilisation.
Patrick Moorhead, founder and chief analyst of Moor Insights and Strategy, was one of the clearest-eyed voices on Terafab when Musk unveiled the project on 22 March — and his analysis aged so well that today's Intel announcement reads almost like a direct response to his publicly stated concerns. On X in March, Moorhead wrote that the proposal was "the most ambitious semiconductor manufacturing bet in history" while simultaneously warning that "Saturday was a presentation, not a fab" — that there were "no ASML, KLA, or Applied Materials orders, no process partner, no manufacturing team." In a LinkedIn article titled "I Was Wrong About Musk on Chip Design. I'm Skeptical About Manufacturing," he had predicted "My best guess: Terafab becomes an advanced technology development centre that partners with Samsung Semiconductor, Intel, or both for the actual high-volume manufacturing on leading-edge nodes." Today, posting on X, Moorhead confirmed the prediction had landed: "This answers the biggest open question from the March 21 announcement: whose process technology?" His post was surgical in its precision — "Intel contributes process technology and advanced packaging. Terafab targets 1 TW/year of compute for AI and robotics. Smart move by Lip-Bu Tan. Intel Foundry needs anchor customers. Musk needs a process partner." On LinkedIn, elaborating further, Moorhead noted that the partnership validates the core thesis that chip design and chip manufacturing are different businesses: "Musk proved he can design competitive silicon. Now he's done the smart thing and partnered with a company that has decades of manufacturing DNA rather than trying to build that capability from zero." He also traced the longer arc: "Back in August 2025, I noted that Tesla's chip strategy appeared to be a Tesla-Intel-Samsung arrangement. Eight months later, that arrangement has formalised into something much bigger." He flagged three open questions that remain: which specific Intel process node Terafab will use, the order timelines for ASML, KLA, and Applied Materials equipment, and how the Intel partnership interacts with Tesla's existing $16.5 billion Samsung deal for AI6 chips through 2033.
The institutional financial community is less uniformly enthusiastic. Bernstein's analysts have estimated that achieving Terafab's stated 1-terawatt annual compute goal would require investment of approximately Rs 420 trillion ($5 trillion). Tom's Hardware's analysis reaches a similar conclusion: at full scale, Terafab would need between 142 and 358 leading-edge logic fabs, each costing $25 to $35 billion to build. ASML shipped only 179 scanners in all of 2025; 12,600 machines would be required at full scale. Morgan Stanley's semiconductor analysts described the effort as "a Herculean task" and estimated total capital expenditure at $35 to $45 billion for the pilot alone, including $20 to $25 billion for wafer fabrication equipment. Bank of America pegged the capital outlay threshold at over $60 billion and concluded that even under optimistic assumptions, Terafab's per-wafer costs would still exceed TSMC's, making the return-on-investment case materially difficult to underwrite.
Barclays analyst Dan Levy, maintaining an Equalweight rating on Tesla, was the most pointed: Terafab could require spending "many multiples" above even his $50 billion bull-case estimate, calling the capital outlook "likely well more than an order of magnitude higher" than Tesla has communicated. Levy noted that Barclays already forecasts Tesla's 2026 free cash flow at negative Rs 252 billion ($3 billion) before any Terafab-related spending. Wedbush's Daniel Ives, Wall Street's most vocal Tesla bull with an Outperform rating and $600 price target on the stock, took the opposite view, calling Terafab "the first step to ultimately what will be Tesla and SpaceX combining forces in a merger likely in 2027" — a prediction that had its own theatrical quality, but one that Wall Street filed without laughing.
None of this makes Terafab useless. It makes Terafab's full-scale vision a long-run aspiration rather than a near-term deliverable. The pilot phase — 100,000 wafer starts per month, Intel-assisted, located on the North Campus of Giga Texas — is a meaningful industrial project. At that scale, it functions as a captive fab for Tesla's and SpaceX's most advanced chips, reducing dependency on TSMC and Samsung, compressing the iteration cycle between chip design and production from months to days, and giving Musk's companies the kind of silicon sovereignty that every major AI player is currently trying to buy its way toward. Applied Materials, Lam Research, and ASML would each benefit directly from a Rs 2.1 trillion ($25 billion) greenfield facility, regardless of its long-term ambitions.
American Chips
The timing of this partnership is inseparable from the political economy of semiconductors in 2026. TSMC's 70 per cent share of the global foundry market, per TrendForce's latest data, represents a geopolitical concentration risk that the U.S. government has spent $52 billion in CHIPS Act funding trying to partially resolve. The risk is simple: Taiwan is 180 kilometres from mainland China. Every AI chip that powers American military systems, autonomous vehicles, financial infrastructure, and national security applications is currently manufactured by a company whose primary fabs sit within range of Chinese ballistic missiles. The semiconductor industry's polite term for this situation is "supply chain concentration." The strategic community's term is somewhat more direct.
Intel, as a CSIS analysis noted in 2024, is no longer a conventional company and can no longer be evaluated as one. It is "a corporate actor on the geopolitical stage." The U.S. government's 9.9 per cent stake in Intel is not a passive investment. It is a statement of national industrial policy: that America requires a domestically owned, domestically operated leading-edge foundry, and that Intel is the only entity currently capable of being that foundry. Terafab, manufactured on American soil by American-supervised Intel processes, fits this narrative so perfectly that it almost seems designed for it — which, given Musk's facility for aligning his ventures with Washington's strategic priorities, it probably was. For countries like India — whose Rs 76,000 crore ($9 billion) Semiconductor Mission is presently focused on mature-node fabrication with TSMC as technology partner — the gap between that ambition and what Terafab represents is the gap between learning to drive and designing the engine.
Who Actually Wins
Here is what actually happened on Tuesday morning, once the press release language is set aside. Intel won a marquee foundry customer at the precise moment it needed one most. Tesla, SpaceX, and xAI won process engineering expertise and manufacturing credibility that no amount of capital can purchase from first principles. The U.S. government won a made-in-America AI chip story at a moment when semiconductor sovereignty has replaced oil security as the defining industrial anxiety of the age. And Terafab — the most-discussed semiconductor project since TSMC broke ground in Arizona in 2021 — received the one thing its original March launch conspicuously lacked: plausibility.
The AI chip market, valued at approximately $61.83 billion in 2025, is projected to reach $84.17 billion in 2026 and $286.7 billion by 2030, per Research and Markets data, growing at a compound annual rate of 36.1 per cent. Nvidia holds approximately 86 per cent of the AI GPU segment. The custom ASIC and edge inference market — precisely the segment Tesla's AI5 and AI6 chips address — is the fastest-growing adjacent category, with Counterpoint Research's Tom Kang noting that "the ramp of 2nm should help Samsung secure higher-value designs, particularly in AI and mobile," though Samsung's 2nm timeline now carries an asterisk the size of Taylor, Texas.
What Intel's post did not disclose is equally significant: no financial terms, no equity structure, no confirmation of whether this is a foundry services contract, a joint venture, a capital co-investment, or some novel arrangement that the semiconductor industry's lawyers are still inventing terminology for. As Bloomberg noted, the announcement provides "limited details about the project's timeline, structure, or specific technical specifications." In Intel's defence, Terafab itself provides limited details about most of those things. The architecture of this partnership will emerge in earnings calls, SEC filings, and the SpaceX S-1 prospectus that drops sometime before the June roadshow. Moorhead, characteristically, put a sharper point on what he is still watching for: specifics on which Intel process node Terafab will use, the order timelines for ASML, KLA, and Applied Materials equipment, and how this partnership interacts with Tesla's existing $16.5 billion Samsung deal. "The partnership is real," he wrote. "The execution timeline is still an open question."
What is already clear, as of Tuesday morning, 7 April 2026, is this: Elon Musk announced last month that he would build a chip factory. Intel announced today that it would build it for him. The rest is engineering. And in semiconductor manufacturing, engineering is the part that takes a decade, requires 300,000 people, and costs five trillion dollars at the scale Musk has in mind.
He has, it should be said, built harder things. Just not faster ones. |
