Key Highlights
- Intel (NASDAQ: INTC) reported second-quarter 2026 revenue of $16.1 billion, up 25%, with Data Center and AI up 59%, growth management calls its strongest in more than fifteen years.
- A $12.5 billion non-cash charge on shares escrowed for the US Department of Commerce drove a GAAP net loss of $11.0 billion, against non-GAAP net income of $2.2 billion and GAAP operating income of $1.8 billion.
- The stock has more than doubled in 2026 but sits roughly 36% below its June high, on a forward P/E near 56 with no usable trailing multiple.
- Foundry still loses more than $2 billion a quarter, and a $293 million external revenue base is too small to change that soon.
Intel (NASDAQ:INTC)'s June-quarter revenue rose 25% as data center demand surged, growth the company calls its strongest in more than fifteen years. A $12.5 billion non-cash charge tied to shares escrowed for Washington turned that quarter into a headline GAAP loss.
A Comeback Narrative Colliding With a Skeptical Market
Intel Corporation (NASDAQ: INTC) has divided opinion in 2026: a company that spent five years losing process leadership, market share and money is now one of the S&P 500's strongest performers, and one of its most volatile. Shares touched a 52-week high of $142.35 and closed at $91.67 at last closing, per aggregated market data: down roughly 36% from that peak, yet still up more than 150% year to date.
The 23 July earnings report brought the argument to a head, containing the best top-line growth Intel has produced since the early 2010s and, on the same page, the largest quarterly net loss in its history. Sorting out which one matters means separating operating performance from an accounting quirk of Washington's equity position.
What Intel Actually Reported on 23 July
For the second quarter of fiscal 2026, ended Intel reported revenue of $16.128 billion, up 25% from $12.859 billion a year earlier. Chief executive Lip-Bu Tan called it "our strongest revenue growth in more than fifteen years."
Data Center and AI revenue rose 59% to $6.3 billion, with segment operating income of $2.474 billion against $633 million a year earlier. The Client Computing and Physical AI Group grew 13% to $8.9 billion. Intel Foundry revenue reached $5.8 billion, up 31%, but the overwhelming majority is internal work: external foundry revenue was $293 million. Foundry's operating loss narrowed to $2.089 billion from $3.168 billion.
Non-GAAP EPS was $0.42, roughly double the consensus of about $0.22 cited in post-earnings coverage. GAAP diluted EPS was negative $2.16.
Margins Recovered; The Bottom Line Did Not
GAAP gross margin expanded to 40.4% from 27.5% and non-GAAP gross margin to 41.8% from 29.7%. GAAP operating income swung to a positive $1.796 billion from a $3.176 billion loss; non-GAAP operating income was $2.770 billion. Chief financial officer Dave Zinsner credited "volume upside driven by higher factory yields and improved cycle times."
The $11.033 billion GAAP net loss comes almost entirely from one line: a $12.529 billion mark-to-market charge on shares escrowed for the Department of Commerce, treated as a derivative liability. Because the liability is marked against Intel's own share price, the stock's rise enlarged the charge. It is non-cash and unrelated to operations, but not cosmetic, since it tracks a real transfer of value to the government. The 10-Q carried the escrowed-share liability at $15.6 billion.
First-half operating cash flow was $8.102 billion versus $2.863 billion a year earlier, against capital expenditure of $6.192 billion. Intel's own adjusted free cash flow measure for the six months was negative $8.419 billion, reflecting $14.2 billion paid to reacquire Apollo's 49% interest in the Ireland Semiconductor Co-Investment Program. Cash was $12.874 billion plus $16.853 billion of short-term investments; total debt was $50.537 billion.
Third-quarter guidance: revenue of $15.8-16.8 billion, GAAP gross margin of 41.0% (42.0% non-GAAP), and GAAP EPS of $0.31 (non-GAAP $0.38).
Where Intel Fits in the AI Compute Build-Out
Intel's AI revenue today is overwhelmingly CPU revenue, not accelerator revenue. Xeon 6 was called one of the fastest-ramping products in company history, and Clearwater Forest, branded Xeon 6+, Intel's first server-class product on 18A, launched during the quarter. Management forecast strong double-digit server CPU unit growth for the industry in 2026 and 2027, and said supply "remains very tight."
The accelerator side is less developed. Crescent Island, an inference-oriented data center GPU on the Xe3P architecture with 160GB of LPDDR5X memory, begins customer sampling in the second half of 2026. Jaguar Shores, an HBM4-based rack-scale design, is expected in 2027 or later. LPDDR5X offers materially less bandwidth than HBM-based rivals, and Gaudi 3 never won meaningful share in large training clusters, the roadmap does not support treating Intel as a near-term NVIDIA challenger in accelerators; the credible case is narrower: inference, edge, and the CPU attached to everyone else's accelerators.
Losing Revenue Share Where the Money Is
Mercury Research data for the first quarter of 2026, reported by Tom's Hardware, put AMD at a record 46.2% of x86 server CPU revenue, up from 41.2% in the fourth quarter of 2025, on 33.2% of units. Intel still ships roughly two-thirds of server CPUs while capturing barely half the revenue. Tan acknowledged Intel is "still behind" in places, pointing to Diamond Rapids and Coral Rapids as the answer. Arm-based custom silicon at the hyperscalers remains a structural drag.
Spending Up, Balance Sheet Reshaped by Outside Capital
Intel raised full-year 2026 capital expenditure guidance to more than $20 billion, roughly $3 billion above prior guidance, and said 2027 spending will be "significantly above" that, concentrated in the US, funded partly by an unusual shareholder register. In August 2025 the US government agreed to buy 433.3 million shares at $20.47 for $8.9 billion, $5.7 billion from remaining CHIPS Act grants and $3.2 billion from the Secure Enclave programme, for a 9.9% passive stake with no board seats, plus a five-year warrant for a further 5% at $20, exercisable only if Intel ceases to own at least 51% of the foundry business. SoftBank signed a $2 billion agreement at $23 per share, and NVIDIA's $5 billion purchase of 214.8 million shares at $23.28 closed, alongside a joint development agreement covering NVIDIA-custom x86 data center CPUs, x86 RTX PC system-on-chips and NVLink integration.
The cost is dilution: 5.043 billion shares were outstanding at quarter-end, with 143 million escrowed shares still undelivered to Commerce.
What Could Move the Story From Here
The clearest near-term catalyst is 14A. Intel has two prospective external customers and expects firm decisions between the second half of 2026 and the first half of 2027, with risk production in late 2027 and volume ramp in 2028; PDK 0.9 is due in October. Meanwhile 18A output ran about 25% ahead of internal targets and Intel halved the cost of its primary Panther Lake SKU year to date. Customer wins are mixed: Microsoft and Amazon Web Services are confirmed 18A customers, while a Wall Street Journal report of a preliminary Apple agreement in May 2026 and a SemiAccurate report on Microsoft's Maia 2 remain unconfirmed.
The Risks That Actually Bite
Foundry still loses more than $2 billion a quarter, and a $293 million external revenue base is too small to change that soon. A 14A customer decision that goes the wrong way would put the post-18A roadmap in question. The escrowed-share derivative will keep injecting large, direction-dependent swings into GAAP results while shares remain undelivered. Intel guided the PC market down by low double digits for 2026, so its largest segment is growing on content and pricing, not units. Debt above $50 billion against a rising capital budget leaves limited slack, and management flagged memory supply constraints, hiring former SK hynix chief Seok-hee Lee to address them.
Valuation Assumes the Recovery Continues
At $91.67 at last closing, Intel's market capitalisation was roughly $462.4 billion and enterprise value about $475.9 billion. Trailing twelve-month EPS was negative $2.30, so no trailing P/E is meaningful, distorted by the escrowed-share charge and a $3.9 billion Mobileye goodwill impairment in the first quarter. Forward P/E was roughly 55.8, price-to-sales about 8.0 and price-to-book about 5.3.
Goldman Sachs maintained a Neutral rating with a $150 price target while lifting estimates by an average of 49%, arguing AMD, NVIDIA and Broadcom offer better revenue visibility and risk/reward. Bank of America warned that AI semiconductor valuations had run ahead of fundamentals.
Conclusion
The operating improvement is real and measurable: gross margin up more than twelve points, a 59% jump in data center revenue, foundry losses down by a third, and the first genuinely competitive 18A server part shipping. What has not happened is the part that decides the outcome: external foundry customers at scale, a competitive accelerator, and an end to server revenue share erosion. Investors are paying a premium multiple for results Intel's own timeline places in 2027 and 2028. The recovery is underway; whether it justifies today's price is a separate question, and the 36% drawdown since June says that debate is live.
This article is for informational purposes only and does not constitute investment advice or a recommendation to buy, sell, or hold any security. Share prices can rise or fall, and past performance does not guarantee future results. Investors should conduct independent research and consult a licensed financial adviser before making investment decisions.






Please wait processing your request...