Intel Raises 2026 Capex Above $20 Billion as CEO Tan Says Demand Now "Substantially" Outpaces Supply Across CPUs, Foundry Wafers and Packaging
Q2 2026 earnings call, July 23, 2026
Intel delivered its seventh consecutive quarter of upside on July 23, but the more consequential news for investors was a sharp increase in capital spending plans and management's insistence that the company is now structurally supply-constrained across nearly every part of its business. CEO Lip-Bu Tan and CFO David Zinsner used the call to reset expectations upward on server CPU demand, 18A and 14A process progress, and the company's nascent ASIC business, while also flagging that PC demand will be softer than normal in the back half of the year.
Capex Guidance Jumps, With 2027 Set to Rise Further
Zinsner told investors Intel is raising 2026 capital expenditures to more than $20 billion, "up significantly versus our expectations entering the year," and that 2027 capex will land "significantly above" the 2026 level. He declined to give a specific 2027 number, saying the company needs "another quarter or two to really solidify" the figure, but the direction was unambiguous. Tooling spend alone will rise 40% in 2026 versus 2025, concentrated in Intel 3, 18A and the new 18A-P variant. Zinsner noted that Intel's total U.S. tool and space capital spending from 2021 through 2026 is approaching $100 billion, which he described as "significantly higher than any other semiconductor company over that time frame."
Management was careful to frame the increase as demand-driven rather than speculative. "This increased investment is a signal of our confidence in customers across all of our business units," Zinsner said, pointing to long-term agreements that now give Intel visibility to forecast demand several years out. Tan added that he will not commit capital "unless I see the yield performance, the IP is ready to serve the customer and also customer engagement." On the balance sheet, Intel holds over $30 billion in cash plus a $10 billion revolver, and Zinsner flagged roughly $10 billion of non-core assets that remain available to monetize if needed, while acknowledging the company "may need to tap the capital markets" if growth continues at this pace.
14A Progress Ahead of Schedule, External Foundry Confidence Building
The most technically significant disclosure was on Intel 14A, where Tan said defect density and transistor performance are "all outpacing 18A development" at the same stage. PDK 0.5 is complete and PDK 0.9 remains on track for October. Intel made the decision in the second quarter to fully commit to high-volume 14A production in 2028, with internal risk production still slated for the second half of 2027. Tan said customer engagement is accelerating now that yield data is becoming visible: "The moment they're starting to see the 0.9 PDK, the yield, they're starting to get excited about what kind of product they want to run on that and how much capacity we can provide them." Notably, external foundry revenue remains small at $293 million for the quarter, and management stopped short of naming any new external customer commitments, leaving the market to wait for concrete proof points beyond internal validation.
18A itself is scaling faster than plan, with output up more than 50% quarter-over-quarter and running roughly 25% above internal targets. Intel has cut the cost of its primary Panther Lake SKU by about 50% year-to-date and expects a further 20% reduction by year-end. 18A-P entered risk production in the quarter, offering a 5% performance uplift while remaining IP-compatible with 18A.
Server CPUs: Strongest Growth in Company History
Data Center and AI Group revenue rose 24% sequentially and 59% year-over-year to $6.3 billion, with operating profit up roughly $1 billion sequentially to a 40% margin. Zinsner called Xeon 6 (Clearwater Forest, the first server product on 18A) "one of the fastest ramping products in Intel history." He singled out Granite Rapids, built on Intel 3, as the tightest-supplied product in the portfolio: "As strong as demand is across all of our products within data center, Granite Rapids is extremely tight because the reception there has been fantastic." Management said its outlook for server CPU unit growth has improved again since the prior quarter, now forecasting strong double-digit unit growth for the industry in both 2026 and 2027, with momentum extending into 2028. Asked to react to a competitor's forecast of a $220 billion CPU TAM by 2030 growing near 45% annually, Zinsner declined to endorse a specific figure but noted the CPU-to-GPU unit ratio is "almost in parity" and could tilt further toward CPUs, an important signal for how Intel is positioning its core franchise against the GPU-led narrative dominating the AI infrastructure conversation.
Client Business Beats on Price, Not Volume
CCPG (the newly renamed Client Computing and Physical AI Group) revenue reached $8.9 billion, up 15% sequentially, but Zinsner was direct that the beat was "largely ASP," driven by mix shift toward higher-end parts and deliberate price increases to offset rising costs. AI PC revenue grew 26% sequentially and now represents two-thirds of client mix. The unit environment is weaker: Intel expects full-year 2026 PC consumption to fall by low double digits, pressured by rising memory prices and constraints, a dynamic it said is consistent with industry peers. Operating profit in the segment fell $173 million sequentially due to inventory charges tied to "stranded" products that no longer made economic sense to complete given component shortages. Zinsner expects client revenue to be roughly flat in the third quarter as Intel deliberately redirects wafer capacity toward higher-margin data center CPUs. "The good news for us is we need it," he said, "because we can't fulfill the demand" on the server side.
ASIC Business Targeting $4 Billion Run Rate
Intel's purpose-built silicon and design services business is becoming a more tangible growth vector. Zinsner disclosed the ASIC business is now running near a $2 billion annualized rate and said Intel expects to reach a $4 billion run rate "in the not-too-distant future," a specific target investors had not previously received. Tan pointed to the new Fortinet security-processor collaboration and existing IPU relationships with hyperscalers as proof points, framing the addressable market at over $100 billion given Intel's combination of CPU/xPU design, IP portfolio and advanced packaging.
Gross Margin Puts and Takes, Memory Ambitions
Non-GAAP gross margin came in at 41.8% in the quarter, roughly 280 basis points above guidance, and Q3 guidance calls for 42%, effectively flat. Zinsner explained that the inventory reserve released from Q2 provides a lift, offset by Panther Lake and Granite Rapids still running below corporate average margins as they ramp. He reiterated that getting gross margin "comfortably into the 40s" every quarter was his top internal goal this year and said the team has achieved that for two straight quarters.
On memory, Tan revealed Intel has hired Seok-Hee Lee, the former CEO of SK Hynix, and is exploring compute-memory integration and stacking technologies as memory becomes what he called a "big supply constraint challenge" and "bottleneck" for AI infrastructure broadly. Details remain thin, but the hire signals Intel sees an opening to re-enter memory-adjacent architecture decisions rather than simply purchasing DRAM and HBM from the three major suppliers.
Zinsner closed with a caution that despite improving supply late in the third quarter and into the fourth, "we will not catch up" to demand in 2026, citing substrate and packaging bottlenecks as the most challenged links in the supply chain, even as front-end wafer output continues to beat internal targets.