One year ago, the United States government made an extraordinary bet on Intel. On Aug. 22, 2025, Washington agreed to invest $8.9 billion in the company, converting previously awarded CHIPS Act and Secure Enclave funding into an equity stake. The goal was bigger than rescuing another struggling American corporation. It was to preserve something the United States had allowed to become dangerously scarce: domestically controlled, leading-edge semiconductor manufacturing.

The government was quickly joined by others. SoftBank invested $2 billion. Nvidia invested $5 billion. Now Intel has priced an upsized $20 billion public stock offering, increased from an originally planned $15 billion after exceptionally strong investor demand. Add it together and Intel has attracted approximately $35.9 billion in capital over the past year. If underwriters exercise their option to purchase another $3 billion in shares, the total will approach $39 billion.

That creates an obvious question: What has Intel done, or what does it intend to do, with all that money?

It is a fair question. It is also one that can produce a misleading answer if we use the wrong scorecard.

If the only measure is Intel Foundry’s external revenue, nearly $36 billion does not appear to have bought very much. Intel Foundry reported $5.8 billion in second-quarter revenue, but only $293 million came from external customers, and much of that reflected Altera’s separation from Intel. The segment recorded an operating loss of approximately $2.1 billion. By that measure, Intel remains a long way from proving that it can build a profitable merchant-foundry business.

But that was never the only assignment, and it may not have been the most important one.

The first and hardest task was to prove that Intel could develop and manufacture leading-edge chips again. On that score, the evidence looks substantially better than it did a year ago.

From Roadmap Promises to Shipping Products

Intel 3, the company’s 3-nanometer-class process, is in high-volume production. Intel 18A, its 2-nanometer-class process, has moved out of development and into commercial products. Intel’s Core Ultra Series 3, code-named Panther Lake, launched at the beginning of this year as the first client platform manufactured on Intel 18A. Intel says the platform is represented in more than 200 PC designs. Xeon 6+, code-named Clearwater Forest, has carried 18A into the data center.

Both products are being manufactured at Intel’s Fab 52 in Chandler, Arizona. The facility is fully operational and represents the most advanced logic manufacturing currently being performed in the United States by an American-owned company. Intel says 18A entered high-volume manufacturing in late 2025, introducing RibbonFET gate-all-around transistors and PowerVia backside power delivery.

These are not literal measurements of 1.8-nanometer transistors. Node names have not corresponded directly to physical transistor dimensions for years. What matters is that Intel 18A belongs in the same leading-edge class of manufacturing technologies that will be used for the next generation of advanced client, data center and AI chips.

A year ago, Intel was still trying to convince the market that 18A would work, that it could achieve acceptable yields and that its factories could manufacture the process at meaningful scale. Intel is no longer relying exclusively on roadmaps and engineering samples to make that argument. It has factories operating, commercial products shipping and customers buying the resulting systems.

The operational evidence is improving as well. Intel said its second-quarter results exceeded guidance in part because higher factory yields and shorter cycle times produced additional volume. Revenue increased 25% year over year to $16.1 billion, while Data Center and AI revenue grew 59% to $6.3 billion. Intel generated $7 billion in operating cash flow during the quarter. Intel’s second-quarter results do not prove that the foundry turnaround is complete, but they do show that better manufacturing execution is producing commercially relevant results.

Futurum Research Director Brendan Burke adds another important piece of evidence. Intel’s 18A output at Fab 52 reportedly came in approximately 25% above its internal second-quarter target and increased more than 50% sequentially. The company has also reportedly reduced the manufacturing cost of its primary Panther Lake SKU by approximately 50% this year, with another 20% reduction targeted.

That is what manufacturing progress looks like. Producing a working wafer is one thing. Producing millions of chips repeatedly, with improving yields, shorter cycle times and falling unit costs, is what begins to turn process technology into a business.

What $20 Billion Can Actually Buy

There is a tendency to treat “general corporate purposes” as evasive language whenever a company raises a large amount of money. Intel did say the approximately $19.7 billion in expected net proceeds from the new offering may be used for capital expenditures, working capital and other general purposes. Intel’s offering announcement does not identify a specific factory, tool purchase or process node.

Capital is fungible, and Intel has not said that every dollar is headed to one project. Even so, the size of the offering gives us a good sense of what it could enable.

Burke estimates that equipping a leading-edge semiconductor fab capable of approximately 40,000 wafer starts per month requires around $25 billion in production machinery. Intel has already spent heavily on buildings, utilities and cleanroom shells. The next wave of spending is increasingly about the lithography systems and other wafer-fabrication equipment that turn those shells into productive factories.

In that context, a $20 billion offering is not an arbitrary pile of money. It is roughly the amount needed to equip one existing Intel fab shell for a leading-edge production ramp. Burke’s Futurum analysis interprets the offering as capital that could prepare Intel for 14A manufacturing, potentially beginning with one of its existing shells.

That is an informed interpretation, not a disclosed Intel allocation. But it provides a tangible answer to the question of where money at this scale goes. It goes into enormously expensive equipment from companies such as ASML, Applied Materials, Lam Research, KLA and Tokyo Electron. It goes into cleanrooms, substrates, process development and advanced packaging. Most of it must be committed years before the resulting chips produce revenue.

The timing becomes more significant when viewed alongside Intel’s changing language about 14A.

Last year, Intel warned that it might pause or discontinue 14A if it could not secure sufficient demand from its own product roadmap and significant external customers. That was a sobering admission. If 18A represented Intel’s attempt to return to the leading edge, 14A would determine whether it could remain there.

Intel’s position is now firmer. In its latest quarterly filing, the company said it committed during the second quarter to completing 14A development. Intel has several future internal products planned for the node, manufacturing expansion projects underway and potential major customers progressing through performance and design milestones. Intel’s Q2 10-Q still makes future capacity dependent on committed demand, as it should, but the possibility of simply abandoning 14A has been replaced by an explicit commitment to finish the process.

Intel 18A is also becoming a process family rather than a one-node demonstration. Intel 18A-P entered risk production in June on the schedule previously provided to customers. Intel claims the enhanced process can deliver 9% higher performance at the same power or 18% lower power at the same performance compared with 18A. It is design-rule compatible with 18A, allowing customers and ecosystem partners to reuse more of their existing design work. Intel’s VLSI Symposium update suggests the company is restoring not only process technology, but the predictable cadence required to sustain it.

America’s Second Source

None of this means Intel has caught TSMC as a commercial foundry. TSMC remains the industry benchmark, with enormous advantages in production volume, yields, design enablement, customer trust and ecosystem maturity. Intel still must demonstrate that it can behave as a neutral manufacturing partner even when potential foundry customers compete with Intel’s own product businesses.

The more supportable conclusion is also the more strategically important one: Intel has become the only plausible American-owned second source for leading-edge semiconductor manufacturing.

TSMC’s Arizona expansion is important and should be welcomed. But TSMC remains a Taiwanese company whose largest and most advanced manufacturing base is concentrated in Taiwan. Intel is the only American-owned company attempting to combine leading-edge process development, advanced domestic wafer manufacturing, packaging, product design and a U.S.-controlled supply chain.

That is what Washington was investing in. The government’s $8.9 billion was funded through $5.7 billion in previously awarded but unpaid CHIPS Act grants and $3.2 billion connected to the Secure Enclave program. It was not entirely new, unrestricted money. It did, however, provide more permanent capital behind a capability with direct economic and national-security implications. The government’s investment agreement explicitly tied the transaction to the expansion of American semiconductor manufacturing.

The SoftBank and Nvidia investments offered another kind of validation. SoftBank said its investment reflected the belief that Intel would play a critical role in expanding advanced semiconductor manufacturing in the United States. Nvidia’s investment came with plans for the companies to develop multiple generations of integrated data center and PC products. It does not mean Nvidia has committed its flagship GPUs to Intel Foundry, but it does mean the world’s most important AI chip company sees strategic value in Intel’s continued strength.

Intel also has real external foundry engagements. AWS entered a multiyear, multibillion-dollar collaboration under which Intel is expected to manufacture a custom AI fabric chip on 18A and a custom Xeon chip on Intel 3. The Intel-AWS agreement is considerably more meaningful than a process evaluation or test wafer. Microsoft has announced an 18A design as well, although the product and expected volume have not been disclosed.

The public market has now provided the largest single piece of Intel’s new capital stack. Intel priced 210.5 million shares at $95 and increased the offering from $15 billion to $20 billion. Futurum, citing Bloomberg, reported that investor orders exceeded $100 billion. Selling stock after Intel’s sharp appreciation allowed the company to raise a remarkable amount of capital with far less dilution than would have been possible a year ago.

That is not charity. Investors are betting that Intel’s improved execution, product demand and strategic position justify financing the next stage.

Now the Business Must Follow

Intel’s progress can be reduced to three questions.

Can Intel develop leading-edge process technology again? Increasingly, yes.

Can it manufacture advanced commercial products at meaningful scale and improving economics? The evidence is getting stronger.

Can it attract enough external customers and production volume to make Intel Foundry profitable? Not yet.

The third question is legitimate, but its answer was always going to trail the first two. Foundry customers choose a process years before a chip enters volume production. They need mature design kits, intellectual property, electronic design automation tools, packaging, predictable yields and confidence that capacity will be available when their products are ready. They also need confidence that Intel will meet its schedules after a decade in which it too often did not.

The first year of Intel’s second chance was about restoring that confidence. Intel brought 18A into high-volume manufacturing, launched client and server products, improved its factory execution, expanded the 18A family and committed to completing 14A. It also demonstrated greater capital discipline by reducing or delaying projects where demand did not justify further spending.

The next year needs to produce a different kind of evidence. Intel must turn evaluations into design wins, design wins into committed wafers and committed wafers into durable external revenue. It must show that its technology recovery can support a commercially sustainable foundry rather than an advanced manufacturing operation serving primarily Intel’s own products.

Nearly $36 billion has not yet bought Intel a successful external foundry business. It may already have bought the United States something it did not have a year ago: a credible, American-owned, 2-nanometer-class manufacturing platform producing commercial chips at scale.

The technology is working. The factories are operating. The products are shipping. The economics are improving, and the capital to prepare for the next node is available.

Now the business must follow.