Intel Blows Through $103 as Tan's Margin Pivot Lands — Nvidia's Stake Now Worth $29.99 Billion

Intel traded at $103.61 on Tuesday, up $7.81 or 8.16%, with 55.087 million shares crossing by late morning against a three-month average of 113.398 million. That volume ranked it the single most actively traded stock in the United States on a session when the Dow Jones Industrial Average (DJI) shed 570.78 points and the S&P 500 (SPX) slipped 0.35% to 7,691.55.

The intraday range ran $96.04 to $103.24, with the print pushing marginally above the session high into the afternoon. Market capitalisation stands at $547.72 billion. The 52-week range spans $24.05 to $142.35, and the stock has appreciated 291.98% across twelve months.

Friday, September 4 set the base. Intel closed at $95.80 after climbing 4.5% on volume of 97.7 million shares. US markets were shut Monday for Labor Day, which concentrated three days of news flow into a single premarket window. The stock opened up 3.9%, topped the S&P 500's premarket gainer list at +3.6%, and then more than doubled that move once cash trading began.

Three separate catalysts landed inside twenty-four hours. Supply-chain reporting indicated a roughly 10% increase to PC processor prices effective October 5. An upgrade to Outperform arrived with a $120 price target. And Intel Foundry disclosed that it has processed more than one million wafers using High NA EUV lithography.

None of those is a small item. Together they address the three questions that have defined the Intel debate for two years: can it price, can it manufacture, and will anyone else use its fabs.

The tape's reaction was one-directional. The stock opened well below its eventual print and was accumulated steadily through the morning, closing the opening gap and extending. That kind of grinding, non-reversing advance in the market's most liquid name is not short covering. It is institutional buying against a supply of stock that has been abundant since the company issued 210 million new shares last month.

Semiconductors broadly participated. Qualcomm (QCOM) rose 4.75% to $176.75. Oracle (ORCL) gained 3.39% to $164.16. Nvidia (NVDA), however, fell 1.46% to $227.00 — the market rotating within AI rather than into it.

The Price Hike: Roughly 10% on PC Processors, October 5

The primary catalyst is straightforward and it goes directly to gross margin. Supply-chain sources indicate Intel plans to raise PC CPU prices by approximately 10% on October 5, 2026.

That would be the third round of increases since the end of 2025. The first landed in the first quarter of 2026 at roughly 10%. The second came in July and covered selected consumer and server parts, with increases ranging from tens of dollars to more than a thousand.

The July round is worth examining because it shows the shape of the strategy. On the consumer side, the Core Ultra 7 270K Plus and Core Ultra 7 250K Plus rose between $30 and $50 depending on model, while the Core Ultra 9 285K held its $599 recommended price and some entry-level parts continued selling below launch. That is not a blanket cost pass-through. It is surgical pricing applied where demand is strongest.

The server side moved far more aggressively. Selected Xeon 6 Granite Rapids parts roughly doubled against mid-2025 retail levels. Certain Xeon 8000 Emerald Rapids models now carry recommended prices above their original launch references, with the largest increases exceeding $1,300. The flagship Xeon 6980P, a 128-core part, moved from $12,460 to $13,955 — a $1,495 increase, or roughly 12%.

Company statements attributed the adjustments to current market conditions, higher supply chain costs and demand exceeding supply for specific models.

The market read is what matters here. A company raising prices three times in twelve months without losing volume is a company with pricing power it did not previously possess. Intel spent the better part of a decade trading price for share against a lower-cost competitor. Pricing into shortage is the opposite behaviour, and it flows directly into the gross margin line that has been the single biggest bear argument on the stock.

Non-GAAP gross margin came in at 41.8% in the June quarter, 280 basis points ahead of guidance, with higher average selling prices from mix and pricing actions explicitly cited as a driver. Another 10% on PC parts compounds that.

Killing the Low-Margin Line Is the Bigger Signal

Buried alongside the pricing report is a strategic item that matters more over a multi-year horizon: chief executive Lip-Bu Tan is reviewing the low-margin Small Core product line, with some products potentially entering end-of-life.

Discontinuing product is a harder decision than raising price. It concedes revenue, market share and design-win presence in exchange for margin. Intel has historically been reluctant to do it, because share in the PC and entry-server market has been the foundation of the ecosystem argument — keep the installed base, keep the software, keep the moat.

Tan is trading that for profitability. Supply-chain sources describe the pricing actions and the product rationalisation as pointing at the same goal: raising overall gross margin and moving away from the strategy of trading price for volume.

The context makes it defensible. Market demand exceeded available product supply in the June quarter due to capacity constraints at Intel's own factories and industry-wide shortages, and the company expects shortages of substrates, memory and other critical components to persist into next year. Full disclosure appears in the quarterly filing at sec.gov.

When supply is the binding constraint, every wafer allocated to a low-margin part is a wafer not allocated to a Xeon selling for $13,955. Rationalising the bottom of the portfolio is not a defensive retreat under those conditions. It is capacity reallocation toward the highest-return end of the mix.

The risk is well understood. Exiting the low end cedes ground to competitors who will take those sockets and may keep them when supply normalises. Intel is betting that the AI compute cycle runs long enough that the trade never has to be reversed.

Guidance already reflects the demand asymmetry. Management expects PC consumption to be sub-seasonal in the second half and down by low double digits for all of 2026 on memory prices and constraints, while the server CPU outlook has improved again, with strong double-digit industry unit growth forecast this year and next and momentum extending into 2028.

One Million Wafers on High NA EUV

The third catalyst is the manufacturing one, and it is the item that speaks to the foundry thesis rather than the product business.

Intel Foundry disclosed at an industry lithography conference that it has processed more than one million wafers using High NA EUV systems. Those machines cost up to $400 million each, and Intel has been the most aggressive early adopter of the technology in the industry.

Separately, the lithography supplier confirmed High NA EUV adoption commitments from Samsung, TSMC and Intel — the three companies capable of building leading-edge logic at scale. Shares of the equipment maker gained 2% on the announcement.

The one-million-wafer figure is the meaningful number. High NA has been dogged by scepticism about throughput, uptime and cost per wafer since the first tools shipped. A million wafers processed is production-scale validation rather than a demonstration, and Intel got there first.

That matters for the foundry pitch specifically. Intel's argument to external customers is that it will have leading-edge capacity inside the United States with process technology at or ahead of the competition. High NA leadership is the most concrete evidence available that the second half of that claim is credible.

It also connects to 18A, the node built on gate-all-around transistors and backside power delivery. Intel has brought 18A to full scale with more than 400 Series 3 designs across consumer and commercial products, with high-volume production running at Fab 52 in Chandler, Arizona. Core Ultra Series 3, code-named Panther Lake, was the first product to ship on the node.

The chain of logic the market is pricing runs: High NA at scale, 18A in high-volume production, 14A on schedule, therefore a credible external foundry offering. Each link has now been demonstrated except the last one, which is the entire remaining question.

Capital spending on tools and clean-room space in the US from 2021 through 2026 is approaching $100 billion, substantially more than any other semiconductor manufacturer has committed domestically.

Technicals: $96.04 to $103.24 and the $142.35 Overhang

The chart has been violent in both directions, which is what a 291.98% twelve-month move looks like from the inside.

Tuesday's action cleared a well-defined level. Intel had been building a staircase of higher lows and higher highs from a base near $85, closing around $95 to $96 across multiple sessions with several green days in the 3% to 5% range. The $95 level carries specific significance: it is the price at which 210,526,315 new shares were issued last month, meaning every buyer in that offering was at breakeven going into Tuesday.

Clearing $95 decisively, then $100, then $103.24, moves the entire offering book into profit. That removes a structural supply overhang — offering participants sitting at a loss are natural sellers into any rally, and they no longer are.

The overhead map from here is sparse. The 52-week high is $142.35, roughly 37% above Tuesday's print, with no dense trading structure between the current level and the high $120s. The stock fell 28% during July and declined 11.22% over the subsequent month, which means the descent through this zone was rapid and left little accumulated volume behind.

Below, the levels are tighter. Tuesday's low at $96.04 is immediate support and coincides with the $95 offering price. Friday's $95.80 close sits inside the same band. A break below $95 would put the offering book back underwater and reopen the mid-$80s where the stock based before the current advance.

Volume is the caveat on the bullish read. At 55.087 million shares against a 113.398 million three-month average, Tuesday's session ran well below normal turnover even at its peak activity. An 8% move on half-average volume is a thin move, and thin moves retrace more easily than heavy ones.

The 52-week low at $24.05 provides context for how far this has travelled. Anyone holding from that level is sitting on a 331% gain, and profit-taking pressure from that cohort has repeatedly capped rallies through 2026.

Relative strength has pushed toward overbought territory on the daily after an 8% single-session move, though it has not reached the extremes that historically preceded sharp reversals in this name.

Q2 by the Numbers: $16.1 Billion and the Fastest Growth Since 2011

The fundamental base underneath the current move is stronger than the stock's reputation suggests.

June-quarter revenue reached $16.1 billion, up 25% year over year and $3.3 billion higher than the same period in 2025. That exceeded guidance by $1.8 billion at the midpoint and represented the fastest revenue growth rate for any quarter since 2011. It was the seventh consecutive quarter of beating financial expectations.

Non-GAAP earnings per share came in at $0.42 against guidance of $0.20 and consensus of roughly $0.21 — a beat of 100% against the company's own forecast. Non-GAAP gross margin printed 41.8%, approximately 280 basis points ahead of guidance, driven by higher revenue, better factory yields and higher average selling prices from mix and pricing actions.

Year-to-date revenue reached $29.7 billion, up $4.2 billion from the first half of 2025. The company ended the quarter with roughly $30 billion in cash and short-term investments, before the August equity raise.

AI-driven businesses grew more than 70% year over year and contributed approximately 70% of total revenue. Purpose-built silicon revenue rose roughly 20% sequentially and nearly tripled year over year. Design services revenue grew nearly threefold.

September-quarter guidance calls for revenue of $15.8 billion to $16.8 billion, a $16.3 billion midpoint, with 42% gross margin, an 11% tax rate and $0.38 in non-GAAP EPS. Consensus entering the print was $15.1 billion and $0.27, so the guide arrived roughly 8% above on revenue and 41% above on earnings.

Non-GAAP operating expenses are being held to roughly $16.5 billion for the year. Non-controlling interest nets to approximately $250 million in each of the third and fourth quarters, rising to approximately $1.1 billion for 2027 and 2028 on a GAAP basis.

Full-year 2025 revenue was $52.9 billion, roughly flat, constrained by industry-wide supply. The 2026 trajectory — $29.7 billion through June plus a $16.3 billion guided third quarter — implies annual revenue in the low-to-mid $60 billion range, a step-change rather than a recovery.

Releases are posted at intc.com.

Data Center at $6.3 Billion and Up 59% Is the Real Engine

Strip the segments apart and one line explains almost all of the growth.

Data center and AI revenue rose 59% year over year to $6.3 billion in the June quarter, and 40% across the first half. Client computing revenue rose 13% year over year and 7% year to date. Intel Products revenue overall increased 28% from the prior-year quarter.

The increases came primarily from average selling price gains, the majority driven by a higher mix of premium products sold, with demand-based pricing actions contributing to a lesser extent to offset higher input costs.

That sentence is the whole margin story in company language. Intel is selling more expensive parts and charging more for them, and both effects are working simultaneously.

Xeon 6 has been one of the fastest-ramping products in company history, with year-over-year server growth the strongest on record. Management stated that data center operations cannot keep up with orders, leaving the company unable to fully meet customer demand.

The most durable detail is contractual. Intel has signed ten long-term agreements with server CPU buyers, structured around either fixed pricing commitments or guaranteed purchase volumes. Those contracts convert a cyclical business into something closer to a subscription, and they materially reduce the risk that the current pricing environment reverses when supply normalises.

The demand thesis rests on a specific architectural argument: as AI workloads shift from training toward inference and agentic systems, demand for general-purpose server CPUs rises rather than falls. Training is GPU-dominated. Inference and agentic orchestration require substantially more CPU per unit of accelerator, which is why the ratio of CPUs to GPUs in deployed systems is the number bulls point to.

If that shift is real, Intel's data center franchise has a multi-year runway that has nothing to do with recapturing share from its direct x86 competitor. If it is not, the current growth rate is a supply-shortage artifact.

The company forecasts strong double-digit industry unit growth in server CPUs this year and next, with momentum extending into 2028.

The Foundry Problem: $5.8 Billion In, $293 Million External

The bull case has one hole, and it is large.

Intel Foundry generated $5.8 billion in June-quarter revenue, up 31% year over year. External foundry revenue within that figure totalled $293 million. The overwhelming majority of Intel Foundry's revenue comes from manufacturing chips for Intel.

The segment posted a $2.1 billion operating loss in the quarter.

That is the arithmetic that has kept a large portion of the market sceptical regardless of how well the product business performs. A foundry serving primarily internal demand is a cost centre with a revenue line attached, and a $2.1 billion quarterly loss against $293 million of genuine external business is not a business — it is a bet.

The strategic history reflects how expensive that bet has been. Intel launched IDM 2.0 in March 2021, creating a foundry services arm and committing roughly $20 billion to two Arizona fabs while pursuing a five-node roadmap culminating in 18A. It relaunched the operation as Intel Foundry in February 2024 and won named 18A commitments from Microsoft and AWS.

The expansion has since been scaled back substantially. Planned fabs in Germany and Poland were cancelled. Ohio construction was slowed. Management initially indicated 14A could be paused without a major external customer before committing in 2026 to complete the node's development.

The central unresolved question has not changed: whether Intel can secure a significant disclosed outside customer for 14A and convert foundry capacity into external business.

Early signals exist. SK Hynix is reportedly evaluating Intel Foundry for HBM4E base dies — advanced packaging and base-die work rather than leading-edge logic, but a real external workload from a memory leader. Design services revenue growing nearly threefold points in the same direction.

Capital expenditure guidance for 2026 was raised to more than $20 billion, with 2027 forecast significantly above that, the vast majority spent across the US network.

18A Is Shipping, 14A Is the Entire Thesis

The node roadmap is where the next three years get decided.

Intel 18A is in high-volume production at Fab 52 in Chandler, Arizona. It is the first Intel node with gate-all-around transistors and backside power delivery, and more than 400 Series 3 designs are in the market across consumer and commercial products. Panther Lake, marketed as Core Ultra Series 3, was the first 18A product, with the initial SKU shipping before the end of 2025 and broad availability from January 2026.

That is execution. Two years ago the market questioned whether Intel could deliver a leading-edge node at all.

Intel 14A is the node that determines whether the foundry becomes a business. Risk production is planned for 2027 with a high-volume ramp committed for 2028. It is the first node designed from the outset to serve external customers as well as internal products, and management has stated that 14A is ahead of where older technologies stood at the equivalent point in their cycles.

Landing a marquee external customer on 14A is the single most important catalyst available to this company. Reports have circulated that Nvidia is evaluating Intel's foundry for its 2028 Feynman generation. If that materialises, it would validate the manufacturing roadmap more decisively than any other single data point, because Nvidia currently has no reason to use a second source except capacity and geography.

The $20 billion raised in August exists specifically to fund the gap between now and 2028. Capital expenditure exceeding $20 billion in 2026 and higher again in 2027, against a foundry losing $2.1 billion a quarter, requires a balance sheet that can absorb several more years of negative segment economics.

The pricing actions and the low-margin product rationalisation feed the same equation. Every basis point of gross margin improvement in the product business funds another quarter of foundry investment without additional dilution.

That is the structural link between Tuesday's price-hike headline and the long-term thesis, and it is why an 8% move on a pricing report is not an overreaction.

$20 Billion at $95 and Exactly What It Bought

The August capital raise is the most consequential financial event in Intel's recent history, and the details matter.

Intel announced a $15 billion underwritten public offering of common stock on August 10, then upsized it the same day to $20 billion on demand, pricing 210,526,315 shares at $95 per share. Underwriters received a 30-day option to purchase up to 31,578,947 additional shares at the offering price less discounts. The offering closed August 12. Net proceeds were approximately $19.7 billion assuming the option went unexercised. The pricing release is filed at sec.gov.

Stated use of proceeds was general corporate purposes including capital expenditures and working capital. The stated rationale was customer signalling of a strong and sustainable demand environment driven by AI compute investment, with physical AI, purpose-built silicon, advanced packaging and external wafers identified as growth areas.

A second offering of roughly $15 billion plus a $2.25 billion overallotment is planned, targeting AI compute and advanced packaging capacity while preserving an investment-grade credit profile.

Dilution runs 4% to 5% on earnings per share. Against a market capitalisation of $547.72 billion and approximately 5.29 billion shares outstanding, the 210.5 million new shares represent roughly 4% of the register.

The market's reaction to an upsizing is the tell. A company that announces $15 billion and prices $20 billion because demand exceeded supply is not a company struggling to fund itself. Institutional appetite at $95 was deep enough to absorb a third more paper than originally offered.

The $95 offering price also carries an uncomfortable comparison. It is more than four times the $20.47 per share at which the US government took its position when CHIPS Act support was partly converted into equity.

Sell-side reaction split. Targets range from $85 at the low end to $145 at the high end, with the mean sitting above $121 — well above the current price. One firm called the raise a net positive while explicitly flagging the 4% to 5% dilution. Two others trimmed targets to $112 and $92 on valuation compression while acknowledging the raise clears a key overhang. Tuesday's upgrade carried a $120 target citing turnaround momentum, server CPU shortages and a foundry collaboration.

The Shareholder Register: Nvidia, Washington and SoftBank

Intel's ownership structure is unlike any other large-cap semiconductor company, and it functions as a floor under the stock.

Nvidia agreed to purchase $5 billion of Intel common stock at $23.28 per share in September 2025, alongside a collaboration to build Nvidia-custom x86 CPUs for data center platforms and x86 system-on-chips with Nvidia RTX GPU chiplets for PCs, connected over NVLink. That position is now disclosed at roughly $29.99 billion.

The appreciation is the point. Nvidia paid $23.28 and the stock trades at $103.61 — a 345% gain on a $5 billion commitment, now worth close to $30 billion. That is not a strategic stake any more. It is one of the largest single equity positions in the semiconductor industry, held by the company with the most credibility in AI.

The US government holds a passive stake of approximately 9.9%, converted from federal CHIPS support at an effective $20.47 per share. SoftBank also invested. Tiger Global raised its position during the June quarter.

That register creates an unusual dynamic. Roughly a fifth of the company is held by parties with strategic rather than financial motives, none of whom is a natural seller into strength. It absorbs supply — including the 210 million shares issued in August — and it removes a category of downside pressure that ordinarily accompanies heavy dilution.

It cuts the other way on governance. A passive government stake in a company simultaneously receiving federal manufacturing support creates a policy dependency that is not present at any competitor, and policy can change.

The strategic value is more concrete than the ownership percentages suggest. Nvidia's stake and the accompanying co-development agreement give Intel access to the AI ecosystem it failed to build on its own with Gaudi. The x86-plus-RTX chiplet architecture for PCs and the custom x86 CPUs for Nvidia's data center platforms are real product lines with real volume attached, independent of whether Nvidia ever becomes a foundry customer.

Under Lip-Bu Tan, the company has stabilised, recruited those partners, shipped its first 18A products and secured billions in state support. The register is the evidence.

Valuation: $547.72 Billion Against Roughly $64 Billion of Revenue

The valuation math is where the bull and bear cases stop overlapping.

At $103.61, Intel carries a $547.72 billion market capitalisation. Trailing GAAP profitability does not support a meaningful price-to-earnings ratio — the March quarter posted a GAAP loss of $0.73 per share against non-GAAP earnings of $0.29.

Work from revenue instead. First-half 2026 revenue was $29.7 billion. Third-quarter guidance midpoints at $16.3 billion. A fourth quarter at similar run-rate puts full-year revenue near $62 billion to $65 billion. Against $547.72 billion of market value, that is roughly 8.5 times sales.

For context, that multiple sits well above where Intel traded for most of the past decade and reflects a market pricing a manufacturing turnaround rather than a chip company's current earnings.

On non-GAAP earnings the picture is stretched further. Second-quarter EPS of $0.42 and third-quarter guidance of $0.38 annualise to roughly $1.60. At $103.61, that is approximately 65 times forward non-GAAP earnings, before applying the 4% to 5% dilution from the August raise.

Sixty-five times forward earnings for a company with a segment losing $2.1 billion a quarter is a demanding number by any conventional standard.

The counterargument runs through the margin trajectory. Gross margin moved from the 37.9% area at the end of 2025 to 41.8% in the June quarter, with 42% guided for September. Three rounds of price increases inside twelve months, plus rationalisation of the lowest-margin products, plus a mix shift toward $13,955 server parts, points toward a materially higher structural margin. On a normalised gross margin in the high 40s or above, the earnings base looks entirely different.

Operating leverage compounds it. Non-GAAP operating expenses are held to roughly $16.5 billion for the year against revenue growing 25%. Every incremental dollar of revenue at 42% gross margin drops substantially to operating income.

The mean sell-side target above $121 implies roughly 17% upside from the current print. The range from $85 to $145 spans a 41-point spread, which is an honest reflection of how binary the foundry outcome is.

Competition: Nvidia, AMD, TSMC and the Memory Squeeze

Intel competes on three fronts simultaneously, and its position differs on each.

Against Nvidia, it is no longer a direct competitor in accelerators — it is a partner and a supplier. Nvidia fell 1.46% to $227.00 on Tuesday against a $5.481 trillion market capitalisation and a 29.13 trailing P/E, while Intel rose 8.16%. That divergence on a single session captures the rotation: capital moving down the AI stack toward the components and capacity where supply is tightest.

Against AMD in x86, the fight is over share in servers and clients. AMD gained in the premarket alongside Intel, indicating the market is treating server CPU scarcity as an industry condition rather than a share shift. Both companies are reportedly raising server CPU prices in constrained markets, which is the behaviour of a duopoly with more demand than capacity rather than two firms fighting over a fixed pool.

Against TSMC in foundry, Intel is the distant challenger. TSMC plans up to $64 billion of capital expenditure in 2026 against Intel's more than $20 billion. Intel's counter is geography and technology — leading-edge capacity inside the United States, with High NA EUV deployed at scale ahead of everyone else. Both TSMC and Samsung have now committed to High NA adoption, which narrows that lead over time.

The binding constraint across all three fronts is not competition. It is supply. Industry-wide shortages of substrates, memory and other critical components are expected to persist into next year and may limit Intel's ability to fully meet customer demand.

Memory is the specific problem for the client business. Rising memory prices and constraints are the stated reason PC consumption is forecast to run sub-seasonal in the second half and down by low double digits for all of 2026. Micron (MU) held $1,016.15 on Tuesday and Sandisk (SNDK) was quoted at $1,740.00, up 11.90%, which shows where the pricing power in that part of the chain currently sits.

Intel's response — raise prices, drop low-margin parts, allocate capacity to servers — is the rational one when the constraint is wafers rather than orders.

Verdict: Margin Expansion Is Real, the Foundry Bet Is Not Yet Settled

Intel at $103.61, up 8.16% on 55.087 million shares, is being repriced for a margin story rather than a foundry story, and that distinction matters because the margin story is already visible in the numbers while the foundry story is still a 2028 event. The evidence on pricing is now unambiguous: a third round of increases arriving October 5 at roughly 10% on PC parts, following a first-quarter round of similar size and a July round that pushed the flagship Xeon 6980P from $12,460 to $13,955, alongside a review of the low-margin Small Core line for end-of-life. That produced 41.8% non-GAAP gross margin in the June quarter, 280 basis points above guidance, with 42% guided for September, on revenue of $16.1 billion that grew 25% and beat guidance by $1.8 billion — the fastest growth since 2011 and the seventh consecutive beat, with data center up 59% to $6.3 billion and ten long-term server contracts locking fixed pricing or guaranteed volumes. Against that sits Intel Foundry: $5.8 billion of revenue, $293 million of it external, and a $2.1 billion quarterly operating loss, funded by a $20 billion raise priced at $95 that diluted holders 4% to 5% and a second $15 billion deal still to come. Technically, clearing $95 moved all 210,526,315 offering shares into profit and removed the most obvious supply overhang, leaving thin structure between here and the high $120s with the 52-week high at $142.35 above and $96.04 to $95 as the line that reopens the mid-$80s below. At roughly 8.5 times 2026 revenue and 65 times forward non-GAAP earnings before dilution, the current price already assumes gross margin keeps climbing toward the high 40s and that 14A lands a disclosed external customer before its 2028 ramp. The mean target above $121 implies 17% from here and the $85-to-$145 spread is an honest measure of how binary that is. Setup is constructive above $95 and structurally unresolved until 14A has a name attached to it — and Tuesday's volume, at less than half the three-month average, argues this move needs confirmation before the $120 handle is anything more than a target on paper.

That's TradingNEWS