In Q2 2025, Intel reported $12.9 billion in revenue, which was essentially flat year over year, with a GAAP loss per share of $0.67 driven by $1.9 billion in restructuring charges and about $1.0 billion of non‑cash items.
Management guided Q3 revenue between $12.6 and $13.6 billion and reiterated $18 billion of 2025 capex, which is a snapshot that looks grim on the surface.
Intel’s “Intel Products” businesses, meaning PC and Data Center & AI, earned $2.7 billion in operating income in Q2 while Foundry posted a loss of $3.17 billion operating loss for the quarter and just $22 million of external revenue.
Those two facts alone explain the consolidated loss and the depressed multiple. Few investors have internalized how much the foundry losses are masking the profitability of the PC and server franchises right now.
Of late, Intel carried $50.8 billion of total debt against $21.2 billion in cash and short‑term investments, which really is solid liquidity but not a balance sheet you want to carry indefinitely through a foundry build‑out. The dividend has been paused this year, a painful but rational choice given those priorities.
Bigger Optionality at Foundry
If the foundry is the problem, it’s also the prize. Intel secured up to $7.865 billion in direct U.S. CHIPS Act funding plus up to $11 billion in loans, which are real, signed incentives that directly offset the cost of fabs in Arizona, New Mexico, Ohio, and Oregon.
This is cash and financing that lower Intel’s effective cost of capital for manufacturing. On the demand side, Intel isn’t just “hoping” someone shows up. Microsoft publicly said it will design a chip for Intel’s 18A process, and Intel has quantified more than $15 billion in expected lifetime deal value for its foundry business across 18A wafer and advanced packaging.
Technically, Intel also planted a flag where its rivals haven’t yet, which is High‑NA EUV. Intel received and assembled the industry’s first High‑NA systems in Oregon, and ASML confirmed shipment of the first EXE:5200B high‑volume tool this summer, key to production nodes beyond 18A later in the decade.
Being first with High‑NA doesn’t guarantee wins, but it does give Intel a credible route to pattern smaller features sooner, and that matters to advanced customers choosing a long‑term partner.
Levers Most Miss
Intel’s “Smart Capital” deals effectively convert parts of fab projects into joint ventures with infrastructure investors, lowering Intel’s upfront cash burden and spreading risk.
In 2024, Apollo invested $11 billion for 49% of Fab 34 in Ireland and Intel retains control and output rights. Rating agencies highlighted how these SCIP structures are equity‑like for Intel even though they do carry minimum‑production obligations. These details are buried in filings that rarely make it into headline models.
There are portfolio moves, too. Intel agreed in April to sell 51% of Altera (the FPGA business) to Silver Lake at an $8.75 billion valuation, deconsolidating a money‑losing segment and freeing up cash as the deal closes in the second half.
And in July, Intel sold 57.5 million Mobileye shares at $16.50, plus a concurrent Mobileye repurchase, while converting another 50 million Class B shares.
Intel remains the majority owner, but it unlocked close to $1 billion in liquidity. Those transactions are factual balance‑sheet upgrades that extend runway for the foundry plan.
One more obscure nugget is the Arizona foundry joint venture with Brookfield, which includes minimum wafer‑start commitments, if Intel under‑utilizes, it owes damages.
Intel Is Doing a Lot Well
NVIDIA’s DGX B300, a Blackwell Ultra system designed for frontier‑model training, uses Intel’s Xeon 6 (6776P) as the host CPU.
That proves something important, which is in the AI compute stack, CPUs still matter for orchestration, memory capacity, and feeding GPUs, and Intel continues to win those sockets. NVIDIA
On accelerators, Intel’s Gaudi 3 is a price‑performance play aimed at inference and some training workloads. Volume PCIe availability is slated for the second half of 2025 through partners like Dell, and Intel has showcased third‑party tests where Gaudi 3 instances undercut H100 economics in cloud settings.
Even if you haircut those claims, a credible, lower‑cost alternative earns a seat at the table for enterprises that care about TCO more than absolute peak speed.
Back in the CPU lane, the Xeon 6 family (efficiency‑core Sierra and performance‑core Granite) is finally a modern, power‑aware answer to AMD’s momentum, Intel’s own data center segment turned in a profit in Q2, and you can see why large fleets considering “AI‑adjacent” refreshes are kicking Intel tires again. The product engine is not fixed, but it’s not sputtering either.
Quick Valuation Snapshot
Intel’s market cap sits near $95–96 billion. Netting $50.8B of debt against $21.2B of cash gets you an enterprise value around $125 billion.
Annualize the Q2 operating income from “Intel Products” ($2.7B × 4 = $10.8B) and if you ascribe a 10–12× EBIT multiple to the PC+server engine alone, a discount to higher‑growth peers, but not crazy for a cyclical duopoly, you’re in the $108–$130 billion range, before giving the foundry any option value, and while the foundry’s losses depress consolidated results.
That back‑of‑the‑envelope math doesn’t prove a bargain, but it does suggest the stock already prices the foundry near zero (or worse) and gives little credit for CHIPS incentives, High‑NA leadership, or marquee customers like Microsoft.
What Will Work and Fail?
The bear case is foundry losses persist longer than expected and 18A slips, leaving external customers to hesitate.
Gaudi 3 struggles to win share in an Nvidia‑centric market and ARM‑based PCs nibble at x86 volumes.
Intel’s workforce was reduced to roughly 75,000 in Q2, thanks to a pause on some European projects while it consolidates assembly into Vietnam and Malaysia.
The bull case is less cinematic but would involve Intel stabilizing PC and CPU share with a competent Xeon/Lunar Lake/Arrow Lake cadence, and winning host CPU sockets in AI systems, as well as converting one or two blue‑chip foundry customers into meaningful 18A wafers as High‑NA tools come online.
Add in the CHIPS cash, the Apollo co‑investment, Altera deconsolidation, and Mobileye monetization, and you’ve got a path where the financial optics improve before the full foundry flywheel spins.
Is Intel Stock Undervalued?
On a one‑year view, this is a self‑help story with headline risk. On a three‑ to five‑year view, there’s a reasonable case that today’s price undervalues (1) the earnings power of Intel’s existing products once foundry losses narrow, and (2) the option value of a subsidized, technologically differentiated foundry at precisely the moment the West wants a second TSMC.
The market is valuing Intel as if the foundry is a bottomless pit. The footnotes, and a few beneath-the-surface wins, say it might be a springboard. Patient investors who can live with volatility may find that the reward‑to‑risk finally tilts in their favor.
The author has no position in any of the stocks mentioned. Financhill has a disclosure policy. This post may contain affiliate links or links from our sponsors.