Intel’s stock surged more than 7% in a single session last week, propelled by two seemingly unrelated developments that, taken together, signal a dramatic reconfiguration of who builds advanced chips in the United States and why. The first: Intel’s formal unveiling of what it calls the Terafab, a sprawling semiconductor manufacturing complex in New Albany, Ohio, that the company claims will become the largest AI chip fabrication site on the planet. The second: reports that Elon Musk’s artificial intelligence company, xAI, is exploring a massive custom chip order that could funnel billions of dollars into American foundry capacity. The convergence of these two stories isn’t coincidental. It’s the clearest sign yet that the race to secure domestic chip production has entered a new, more urgent phase.
As Business Insider reported, Intel CEO Lip-Bu Tan personally led the announcement of the Terafab project during a high-profile event in Ohio, framing it as the centerpiece of Intel’s turnaround strategy. The facility, backed by billions in federal CHIPS Act subsidies and state incentives, is designed to manufacture chips at the most advanced process nodes — the 18A and subsequent generations that Intel hopes will make it competitive with Taiwan Semiconductor Manufacturing Company. Tan called it “the most advanced semiconductor manufacturing facility in the world,” a bold claim for a company that has spent the better part of five years watching TSMC and Samsung eat its lunch in contract manufacturing.
The Terafab isn’t just big. It’s enormous. Intel has committed roughly $28 billion to the Ohio site alone, with construction already underway on multiple fab buildings. When fully operational, the complex is expected to house several fabrication plants, packaging facilities, and research labs — all co-located to reduce the logistical friction that plagues chip production when different steps happen on different continents. The sheer scale is Intel’s bet that size and integration can offset the lead TSMC has built over the past decade.
But a factory is only as valuable as the customers willing to fill it.
That’s where Musk enters the picture. According to multiple reports, xAI has been in discussions about commissioning a custom AI accelerator chip — one that would be manufactured in the United States rather than at TSMC’s facilities in Taiwan. The logic tracks with Musk’s broader posture on supply chain risk. He’s spoken publicly about the geopolitical vulnerability of concentrating advanced chip production in Taiwan, a position that has gained traction in Washington and on Wall Street alike. If xAI were to commit to a major order with Intel Foundry Services, it would represent exactly the kind of anchor customer Intel desperately needs to prove its foundry business model works.
The financial stakes are staggering. Custom AI accelerators for a company operating at xAI’s scale — Musk has spoken of building clusters with 100,000 or more GPUs — could translate into a multibillion-dollar manufacturing contract over several years. For Intel, which reported foundry revenue of just $4.7 billion in 2024, a single customer of that magnitude could fundamentally alter the unit economics of the Terafab and Intel Foundry Services as a whole.
Wall Street noticed. Intel shares jumped on the combined news, with analysts at several firms upgrading their outlook on the stock. The enthusiasm, however, comes with heavy caveats. Intel’s 18A process node, which is the technology it’s pitching to potential foundry customers, hasn’t yet entered volume production. TSMC’s equivalent nodes are already shipping to Apple, Nvidia, and AMD. Intel is essentially asking customers to place enormous bets on a manufacturing process that remains, for now, a promise rather than a proven product.
And promises have burned Intel before.
The company’s previous CEO, Pat Gelsinger, launched the IDM 2.0 strategy in 2021 with similarly grand ambitions for Intel’s foundry business. Under Gelsinger, Intel secured the initial CHIPS Act funding, broke ground in Ohio, and signed early agreements with companies like Microsoft. But execution stumbled. The 20A and 18A nodes fell behind schedule. Key engineering talent departed. And the foundry business continued to hemorrhage cash, posting operating losses that made investors increasingly skeptical about the entire venture. Gelsinger’s departure in late 2024 was, in part, a consequence of that skepticism.
Lip-Bu Tan, who took the reins in early 2025, has adopted a different tone. Less evangelical, more operational. His background running Cadence Design Systems — one of the premier electronic design automation firms — gives him credibility with the engineering community that Gelsinger, for all his charisma, had begun to lose. Tan has been methodical about resetting expectations while simultaneously accelerating the timelines that matter most: getting 18A into customers’ hands and proving it can yield competitive chips at scale.
The Terafab announcement is Tan’s first major public statement of intent. It’s also a political document. The event in Ohio featured appearances by state and federal officials, a reminder that Intel’s manufacturing expansion is as much a government-backed industrial policy project as it is a corporate strategy. The CHIPS and Science Act, signed into law in 2022, allocated $52.7 billion for domestic semiconductor manufacturing and research. Intel has been the single largest beneficiary, with commitments totaling roughly $8.5 billion in direct subsidies plus billions more in loans and tax credits.
That government backing creates a peculiar dynamic. Intel’s foundry business doesn’t need to be immediately profitable — it needs to be credible enough to attract customers and keep the federal money flowing. The Musk connection, whether or not a formal deal materializes, serves that purpose beautifully. It signals demand. It signals relevance. And it gives Intel a narrative that extends beyond its own internal chip designs, which have been losing market share to AMD and Arm-based competitors for years.
Musk’s interest in domestic chip production also reflects a broader shift among major AI companies. The concentration of advanced manufacturing in Taiwan has become a boardroom-level concern at virtually every large technology firm. Nvidia, which designs the GPUs that dominate AI training, manufactures almost exclusively at TSMC. So does AMD. Apple’s entire chip line — the M-series and A-series processors — comes from TSMC fabs. A Chinese blockade or invasion of Taiwan, however unlikely, would cripple the global technology industry overnight. That risk calculus is pushing companies to diversify, and Intel is positioning itself as the primary American alternative.
TSMC, for its part, is building its own U.S. fabs in Arizona, with the first facility expected to begin production in 2025. But TSMC’s American operations will be a fraction of its Taiwan capacity for years to come, and the company has faced well-documented challenges adapting its management culture to the U.S. labor market. Samsung is also expanding in Texas, though its foundry business has struggled with yield issues that have cost it customers. Intel’s pitch is that as an American company with decades of domestic manufacturing experience, it can scale U.S. production faster and more reliably than its Asian competitors.
Whether that pitch holds up depends almost entirely on 18A.
Intel has said it expects to begin risk production on 18A in the second half of 2025, with volume production following in 2026. Early test results, according to the company, have been encouraging — Intel claims the node is meeting or exceeding internal targets for transistor density, power efficiency, and performance. Independent verification is limited, though some industry analysts who have reviewed Intel’s technical disclosures have expressed cautious optimism. The node uses a combination of RibbonFET transistor architecture and PowerVia backside power delivery, two innovations that, if they work as advertised, could give Intel a genuine technical edge over TSMC’s N2 node.
That’s a big if. Semiconductor manufacturing at these dimensions — we’re talking about features measured in angstroms, fractions of a nanometer — is among the most complex industrial processes humans have ever attempted. A single defect in a photolithography step can render an entire wafer worthless. Yield rates, the percentage of functional chips produced per wafer, determine whether a process node is commercially viable or a money pit. Intel’s recent track record on yields has been, to put it diplomatically, inconsistent.
The xAI angle adds another layer of complexity. Custom chip design is a multiyear process. Even if Musk’s team signed a deal with Intel tomorrow, the resulting chip wouldn’t tape out for 18 to 24 months, with volume production following a year or more after that. By then, the competitive picture could look very different. TSMC’s N2 node will be in full production. Samsung may have resolved its yield problems. And new entrants — Rapidus in Japan, for instance — could be offering alternatives.
Still, the symbolism matters. Musk is arguably the most influential figure in technology today, and his public association with Intel’s foundry ambitions lends them a weight that no amount of corporate marketing could achieve. It also creates a potential flywheel effect: if xAI commits, other AI companies may follow, not wanting to be left without access to domestic manufacturing capacity if geopolitical tensions escalate.
There’s a financial dimension to this that deserves scrutiny. Intel’s foundry business is currently structured as a separate reporting unit, a move Gelsinger made to increase transparency and, eventually, attract outside investment. The unit has been posting significant operating losses — more than $7 billion in 2024 alone. Tan has indicated he’s open to bringing in strategic investors or even taking the foundry unit public at some point, but only after it demonstrates commercial viability. A marquee customer like xAI would be a critical proof point.
The broader semiconductor industry is watching all of this with a mixture of fascination and wariness. Intel’s competitors have spent years capitalizing on its missteps, and they’re not eager to see it reemerge as a credible foundry player. TSMC, in particular, has built its business on being a neutral manufacturer — it doesn’t compete with its customers by designing its own chips. Intel, which still designs and sells its own processors, faces an inherent conflict of interest that has made some potential foundry customers uneasy. Tan has tried to address this by erecting organizational firewalls between Intel’s design and manufacturing arms, but trust takes time to build.
For investors, the calculus is straightforward but uncomfortable. Intel’s stock, which traded above $60 in early 2024, has been battered — hovering in the low $20s for much of early 2025 before the recent pop. The Terafab and xAI news provided a jolt, but the stock remains far below its highs. The bull case rests on 18A working, the foundry business attracting major customers, and Intel’s own chip designs regaining competitiveness. The bear case is that 18A disappoints, the foundry losses continue to mount, and Intel burns through its CHIPS Act subsidies without achieving self-sustaining profitability.
The truth will probably land somewhere in between. Intel has real technical talent, real manufacturing infrastructure, and real government support. It also has real execution problems, real competitive disadvantages, and a real credibility gap that one flashy announcement can’t close. The Terafab is a massive physical commitment — concrete and steel and billions of dollars poured into the Ohio ground. It will take years to know whether that commitment was visionary or foolish.
Musk, meanwhile, has his own track record of making bold pronouncements that don’t always materialize on the timeline he suggests. His companies have a pattern of announcing ambitious plans, generating enormous excitement, and then delivering results that are impressive but different from what was originally promised. A custom xAI chip manufactured by Intel would be a remarkable achievement. It would also require sustained focus and investment from a man who is simultaneously running Tesla, SpaceX, xAI, Neuralink, The Boring Company, and the federal government’s cost-cutting initiative.
So the market’s enthusiasm is understandable. But it should be tempered.
What isn’t in doubt is that the United States is in the early stages of a semiconductor manufacturing buildout unlike anything it has attempted in decades. The CHIPS Act money is flowing. The fabs are rising. And the customers — from Musk to Microsoft to the Department of Defense — are signaling that they want American-made chips, even if they cost more and take longer to arrive. Intel has positioned itself at the center of that movement. The Terafab is its largest and most consequential bet. Whether it pays off will depend not on speeches in Ohio or stock pops on Wall Street, but on whether Intel can do the hardest thing in the semiconductor business: make the chips work.
WebProNews is an iEntry Publication