The email went out on a Monday morning, August 10, and by that evening the math had already changed. Intel announced a $15 billion common stock offering to support skyrocketing customer demand for artificial intelligence computing power. Retail shareholders read the word "dilution" and shares of Intel fell as much as 5% following the announcement. Then Tuesday arrived, and the institutional world replied with something closer to a stampede. Demand for the offering exceeded $100 billion, and by Tuesday morning the offering was upsized to $20 billion from the previously announced size of $15 billion. Intel priced the underwritten public offering of 210,526,315 shares at a public offering price of $95 per share, a discount of 6.5% to Friday's closing price, according to Bloomberg calculations. The gap between what retail investors feared and what institutional capital was willing to pay for is the whole story in miniature.
To understand why Wall Street wanted a hundred billion dollars worth of a dilutive semiconductor equity offering in 24 hours, you have to zoom out to the macro. Goldman Sachs's baseline model implies $765 billion in annual AI capital expenditure in 2026, growing to $1.6 trillion in annual capex in 2031. Goldman Sachs is forecasting capex of $1.01 trillion in 2027, rising 32% from its 2026 estimate of $765 billion; overall, the firm is forecasting an astronomical $7.6 trillion in cumulative AI capex from 2026 to 2031. Crucially, analysts have persistently underestimated Big Tech AI spending, with guided 2026 spending now over 150% higher than initial estimates, and actual spending exceeded early forecasts by roughly 25% in 2024, and then exceeded forecasts by a whopping 62% in 2025. In that context, a company that builds the physical substrate of AI issuing stock is not a sign of distress. It is a reservation at the only table that matters.
Intel's specific thesis rests on a technology that most people outside the semiconductor industry cannot picture: advanced packaging. As transistor scaling has slowed, performance gains increasingly come not from shrinking individual chips but from stacking and interconnecting multiple dies inside a single package. Advanced packaging has become a commodity for the tech industry, as significant as semiconductors in today's market, given that manufacturers like Nvidia have resorted to using it to scale performance without entirely relying on Moore's Law. Right now, TSMC entirely sources advanced packaging demand, with supply severely constrained, far more than semiconductors themselves. Intel is positioning itself as the American alternative. In 2026 the latest development is EMIB-T, which adds channels through a bridge to deliver power directly to the chips, improving power efficiency and signal routing required by the latest high-bandwidth memory technologies. On Intel's Q4 2025 earnings call, CFO David Zinsner said early customer engagements suggest that many advanced packaging opportunities could be well north of $1 billion. Reports from Digitimes indicate that while orders from Google and Amazon remain unconfirmed, the direction is commercially logical and aligns closely with Zinsner's view of AI ASIC packaging opportunities. The equity raise funds the factory tooling to make those deals real.
The confluence of the U.S. government's industrial policy and corporate AI capex has also handed Intel a structural backstop its rivals cannot replicate. Intel shares have benefited over the last year from the AI infrastructure build-out and the U.S. government's 10% equity stake aimed at bolstering domestic chip manufacturing. Nvidia put in $5 billion via private placement, and SoftBank contributed $2 billion. Layer a $100 billion order book on top of that, and you have four separate classes of sophisticated capital, sovereign, strategic, financial, and now public-market, all betting the same direction. Last month, Intel posted its fastest revenue growth in nearly 15 years and hiked capital expenditure guidance to $20 billion, citing strong customer demand, and finance chief David Zinsner told CNBC that the company is bracing for a "meaningful increase" in 2027. That forward guidance is precisely why the offering upsized by a third in a single evening: sophisticated buyers were pricing in the 2027 capex cycle, not the 2026 one.
The skeptical case is real and deserves full airtime. With $50.5 billion in debt, negative returns on assets, and a foundry business still years from profitability, the dilution funds a multi-decade capital hole. Intel has accumulated roughly $44 billion of negative free cash flow from 2022 through 2025. Capital spending projects can take years before they generate meaningful returns; semiconductor fabs are particularly expensive, and the economics depend heavily on utilization, manufacturing yields, customer commitments, and the ability to maintain technological competitiveness. Russ Mould, investment director at AJ Bell, cut right to the historical irony: "As a capital-intensive business that went a long way to wrecking its own balance sheet and prospects by focusing on financial engineering rather than physical engineering, courtesy of $82 billion of share buybacks in the 2010s, it makes perfect sense for Intel to raise money, especially after a five-fold increase in the stock price since last August." The worry is not whether the AI capex supercycle is real. The worry is whether Intel can execute fast enough to capture it before the cycle matures and the returns compress. Raising equity on top of a $20 billion capex target says the spending trajectory goes higher still, and that free cash flow will stay negative longer than consensus assumes.
The $20 billion offering is not a vote of confidence in Intel specifically; it is a vote of confidence in the proposition that the physical infrastructure of intelligence is now worth financing the way nations once financed railroads, and whoever owns the factories when the demand peaks owns the returns.