Intel 2025 Q2 Financials
Tan Makes Hard Pivots, Charts a New Course
As we kick off another cycle of earnings reports in the chipmaking industry, as always, we start with Intel INTC 0.00%↑ . One of just three leading-edge chip fabs left in the world, Intel has landed in the most precarious position of all of them, as the combined chip designer and chip fabricator has been battered from both sides. The past couple of years has seen Intel losing chip customers to design rivals like AMD, and even losing business from some of its own internal teams to fab rival TSMC. That is expected to change, but the reality is in a state of transition.
As a result of these struggles, Intel has been in the midst of a restructuring and rebuilding phase for much of the past year as the company seeks a return to profitability. And as they close the books on the second quarter of 2025, those efforts, helmed by new CEO Lip-Bu Tan, are beginning to reach their crescendo.
Alongside today’s earnings release in a letter sent to employees and published on Intel’s website, Tan has announced a series of further initiatives at Intel to restructure the company. Of particular note, Tan has announced a new target for Intel’s already in-progress layoffs, with a goal of trimming another 15% of the company on top of the cuts already made this year. When all is said and done, Intel is targeting a headcount of just 75,000 employees, down significantly from the 96,000 they had at the end of this most recent quarter. Meanwhile on the capital expenditure side of matters, Intel is halting their fab construction plans for Germany and Poland, with an overall goal of consolidating Intel’s manufacturing operations across fewer sites – and more in-line with the actual demand for their fab capacity.
Consequently, while Intel’s had numerous quarters in the red over the last year, Q2’2025 is the most brutal yet, as Intel is taking numerous one-off charges for restructuring, employee buyouts, and accelerated depreciation charges (replacing old equipment with over-purchased new equipment). All of this is ultimately in service of bringing Intel back to profitability in the long run, but in the short run it means a lot of pain for Intel management, employees, and shareholders alike.
Key Takeaways (GAAP)
💵 Q2 Revenue, $12.9b, flat YoY from $12.8b and up 2% QoQ
📈 Q2 Gross Margin at 27.5%, down 7.9pp YoY and down 9.4pp QoQ
💰 Q2 Net Income of -$2.9b, down 81% YoY from -$1.6b and down 355% QoQ
🪙 Q2 EPS -$0.67, down 76% YoY, down 353% QoQ
Highlights
Alongside Intel’s latest restructuring plans, the company has also seen a few notable business changes over the last quarter. Including some product-related news that is likely to be of interest to Intel’s datacenter customers.
SMT is coming back to Intel’s data center processors. Outlined in his letter to employees, Tan is instructing Intel’s DCAI teams to bring back support for simultaneous multi-threading. While still technically present in Intel’s most recent generation of Xeon hardware, this feature was disabled for security reasons. But, citing the need for hyperscale performance, Tan is having it brought back.
Intel has sold $922m in Mobileye shares in the last couple of weeks, shoring up their cash reserves. With 57.5 million shares sold, this represents about 7% of Intel’s total Mobileye holdings – leaving them with approximately 748m shares, and remaining as Mobileye’s largest shareholder. Tan is also expecting the majority sale of Altera to close this upcoming quarter.
Intel has kicked off 18A wafer production in Arizona. With Intel set to ship its first 18A-based product later this year – the mobile-focused Panther Lake – the company’s 18A production capacity is starting to come online. This marks the first non-development fab to undertake 18A production, with the Arizona fabs capable of ultimately processing a far larger number of 18A wafers.
Tan announces the name of the generation after Diamond Rapids. It will be called Coral Rapids, due in the 2028/2029 timeframe.
Financial Overview
For the second quarter of 2025, Intel’s losses have once again accelerated on both a quarterly and yearly basis. At $12.9b in revenue for the quarter, Intel’s revenue is effectively flat on a year-over-year basis. And while this is not the revenue growth that Intel would like to see, it is nonetheless good news for the company, as it’s nearly $500m above the high-end of their guidance for the quarter. So rather than seeing a revenue decline, Intel has been able to hold the line on revenue, a silver lining on an otherwise rough quarter. With most of Intel’s revenue in the client business, normally Q2 is notably higher due to a seasonally weak Q1, but was Intel said last quarter, their Q1 was actually quite good compared to the normal.
Unfortunately, the costs of delivering that revenue for Q2 were even higher than they were in Q1. With everything from a higher cost of sales to restructuring charges, Intel ended the quarter deeper in the red on a GAAP basis, losing some $2.9b.
All told however, Intel is taking more than $2.9b in various one-off charges in Q2. This includes $1.9b in restructuring charges – primarily in the form of corporate downsizing – which is further coupled with $800m in impairment charges for the accelerated depreciation of excess tools with no identified re-use, and finally around $200m in additional one-time period costs. (Intel later clarifies that the $800m impairment was due to overbuying tools for new fabs, and using them to replace older tools in existing fabs, and the old equipment having more value as write-down than as sales).
Otherwise, even excluding those one-off charges, Intel’s non-GAAP results still have the company losing around $441m for Q2. This is a notable dip into the red for the core operations of the company, as non-GAAP net income was positive both in the previous quarter and the year-ago quarter.
The impact on Intel’s gross margins, in turn, has been equally painful. On a GAAP basis, Intel is recording a gross margin of just 27.5%, down nearly 8 percentage points from the year-ago quarter, never mind the company’s halcyon days of 60% margins. And non-GAAP figures aren’t much better here, with Intel recording a non-GAAP gross margin of 29.7% for the quarter. Ahead of Intel’s earnings Q&A, Intel’s written materials on the matter are sparse, so it’s unclear what elements in particular were dogging Intel’s non-GAAP margin for this most recent quarter.
Intel Foundry
💵 Q2 Revenue $4.4b, up 3% YoY
💰 Q2 Operating Loss $3.2b, up from $2.8b loss YoY
📈 Q2 Operating Margin -71.7%, up from -65.4% YoY
The actual chip-making arm of Intel, Intel Foundry continues to find itself in an uncomfortable spot amidst its own turnaround into becoming a major player in the contract foundry world. Complicating matters, new Intel CEO Lip-Bu Tan is setting about in taking Intel’s foundry business in a completely opposite direction from former CEO Pat Gelsinger, with a focus on cautious expansion versus Gelsinger’s plans to pole-vault back into first place on bold technology and rapid capacity expansions to meet the domestic demand for chips. In a lot of respects Intel Foundry’s roadmap remains locked in place due to the years of lead time required to make changes, but Tan is certainly aiming to change as much as he can as quick as he can.
In the meantime, the lack of significant short-run flexibility means that Intel Foundry’s trajectory is essentially unchanged from the past several quarters. Foundry has been operating with a negative operating margin for years now, and the most recent quarter was particularly painful, with an operating margin of -71.7%. The goal is to eventually hit profitability by the end of the decade, but it needs to start going in that direction.
In terms of pure revenue, Intel Foundry booked $4.4b in revenue for Q2, which is a very slight uptick from the year-ago quarter. But with an operating loss of some $3.2b for the business unit, losses have accelerated on both a quarterly and yearly basis.
Overall, Intel is attributing these growing losses to a few different factors. In particular is the write-down of fab assets, which has been baked into Intel Foundry’s operating margins. Coupled with that, as with Q1, the company’s Intel 7 process node remains capacity constrained, meaning that Intel isn’t able to produce as many chips on the older fab node as it would like to. Some analysts asked if that was driving more Intel 4/3 volume, but the answer is no - Raptor Lake on Intel 7 is preferred to Meteor Lake on Intel 3 it seems, as the advanced packaging on Intel 3 is more expensive to run. On the flip side of the coin, Intel is still in the process of ramping up 18A production, all of which comes with its own costs.
Otherwise, Intel Foundry’s technology roadmap remains unchanged, even with Tan’s larger restructuring efforts. 18A remains a critical node for Intel – marking the debut of their GAAFET and BSPDN technologies – albeit with the tacit acknowledgement that Intel hasn’t been able to attract much in the way of external customers for the node thus far. So while 18A (and 18A-P) will be critical nodes for Intel Products – and long-lived ones that will see heavy use well into the next decade – it remains to be seen if and when they’ll have significant adoption by other chip designers. If all goes well for Intel, they hope to be able to sign-up customers for 18A in successive waves after proving its competitiveness, which would have it in contention for n-1 designs that don’t need to be on bleeding-edge process nodes.
Beyond that lies Intel 14A, which is still far enough out for Tan to make some more significant changes to Intel’s roadmap. Tan’s letter to Intel’s employees makes it clear that he is tying the success of 14A far more closely to external customer adoption – meaning that for 14A to succeed, Intel needs to land the external customers that Intel hasn’t been able to attract to date. According to Tan, 14A will be developed “from the ground up in close partnership with large external customers.”
Along those lines, Tan also intends to take a more conservative path with building out 14A capacity. Rather than a “build it and they will come” mentality, Tan intends to only build out as much 14A capacity as Intel Foundry needs. Or as Tan put it:
Going forward, our investment in Intel 14A will be based on confirmed customer commitments. There are no more blank checks. Every investment must make economic sense. We will build what our customers need, when they need it, and earn their trust through consistent execution.
Painting with broad strokes, this leaves in question what will happen with 14A if Intel doesn’t attract sufficient outside use. Does Intel still build out any 14A capacity if Intel Products is the only major user? And can Intel Products alone drive enough usage to make 14A financially viable? We’re still years out from the launch of 14A, but Tan’s conservative approach to building out capacity means that a large capacity build-out for 14A is not guaranteed at this time.
Consequently, along with the Germany and Poland fab plans being put six feet under, Intel is also once again further slowing the build-out of their Ohio fab cluster. With Intel struggling to use all of its (non-Intel 7) capacity at the moment, and with few external customers secured, Tan does not believe that Intel needs additional capacity at this time. And with the current US administration in no rush to hand out CHIPS Act funding to pay for additional fabs, even with the recent AI summit stating that US manufacturing needs to be shored up, the desire for the Ohio fabs has cooled significantly for both sides.
Wafer fabbing aside, Intel’s chip packaging operations are also set to undergo some changes. As part of Tan’s restructuring, Intel’s systems assembly and test (SAT) operations are going to be consolidated, with Intel winding down operations in Costa Rica in favor of their largest sites in Vietnam and Malaysia.
And in the more immediate future, Intel is reducing its revenue expectations for its external advanced packaging services, indicating that demand for Intel’s cutting-edge packaging technologies from external chipmakers is going to be lower than Intel earlier believed. Unfortunately, details here are otherwise slim, but this is in and of itself an interesting development since for the last several quarters advanced packaging capacity has been one of the bottlenecks constraining the production of high-end GPUs and other AI accelerators – all of which need advanced packaging technologies to attach HBM memory to their respective ASICs.
Overall, most of Intel’s long-term guidance today has focused on the foundry business, and it’s no coincidence that “Become a More Financially Disciplined Foundry” is item #1 on Tan’s list of company pillars. Tan believes it’s the foundry business that needs to make some of the company’s biggest changes, and with a remit from the board to take Intel in a very different direction from Pat Gelsinger, he is doing just that.
Data Center and AI (DCAI)
💵 Q2 Revenue $3.9b, up 4% YoY from $3.8b
💰 Q2 Operating Income $633m, up from $242m YoY
📈 Q2 Operating Margin 16.1%, up from 6.4% YoY
The first of Intel Products’ two top-line chip businesses, DCAI has in recent years been the more volatile of Intel’s chip business due to a combination of corporate buying cycles and increasingly stiff competition from AMD and numerous Arm architecture chip designers. In good years, DCAI is the most profitable portion of Intel’s portfolio, so the health of this business unit is critical to Intel’s overall operations.
On a yearly basis, DCAI revenue is just above being flat, with the $3.9b in quarterly revenue marking a 4% increase over Q2’24. Otherwise on a quarterly basis, revenue is down around 5%.
More significant, perhaps, is Intel’s operating income for the business unit, which at $633m is the highest it’s been for the restructured group in at least the last year and a half. With an operating margin of 16.1% it’s still far from Intel’s best days, but this marks the strongest performance from the DCAI group in the last several months. All told, this is the third straight month of operating income growth for the business unit, which Intel is taking as a very positive sign.
The biggest driver for DCAI’s improved fortunes here appears to be Intel’s Xeon 6 product portfolio, which is finally paying some bigger dividends now that it has further ramped. With both Granite Rapids (P-core) and Sierra Forest (E-core) parts shipping in volume, Intel has been able to tap into the AI market by providing host CPUs, most notably for NVIDIA’s DGX B300 boxes – a fact that Intel has been very eager to crow about.
And while applicable to both Intel’s client and DCAI product lineups, it’s notable that as part of his pillars of growth strategy, Lip-Bu Tan is also mandating that every major chip design is reviewed and approved by him before tape-out. Being implemented in the name of discipline, this means Tan is going to have a hand in future DCAI chips – especially Coral Raipids, which will succeed Diamond Rapids and was announced by Tan on today’s earnings call. No specific date was attached to Coral Rapids, but given Intel’s roadmap cadence, we’re presumably looking at a 2028/2029 part.
Finally, on the “AI” portion of DCAI, Tan is making it clear that Intel is done pursuing the high-end training accelerator market, at least for the time being. Instead, Tan is tasking Intel with focusing on inference and emerging AI workloads. This route means Intel won’t be competing head-to-head with NVIDIA in a market they dominate. But it’s a strategy that comes with its own risk as the inference market has a larger number of competitors overall, and emerging workloads are, by definition, an unproven market.
Curiously, throughout all of this there hasn’t been any further mention of the Gaudi architecture. Intel’s AI silicon plans have undergone several changes over the years – most notably with the cancellation of Falcon Shores – and the long-term plan has been to integrate Gaudi into Intel’s broader GPU architecture. It looks like this portion of Intel’s GPU plans remains on-track, at least, as Intel isn’t talking about a future for the Gaudi architecture or IP. Even when mentioning the future of AI, Tan highlighted x86 and the Xe graphics architecture only, not Gaudi.
Client Computing Group (CCG)
💵 Q2 Revenue $7.9b, down 3% YoY from $8.1b
💰 Q2 Operating Income $2.1b, down from $2.6b YoY
📈 Q2 Operating Margin 26.1%, down from 32.4% YoY
The current breadwinner amongst Intel’s business units continues to be the Client Computing Group. While CCG is not enjoying the same kind of operating income growth as DCAI is – operating income is down on both a quarterly and yearly basis – it’s still Intel’s biggest business unit by far, booking $7.9b in revenue for the quarter. Compared to the year-ago quarter, this is down 3%, though it is up slightly from Q1’25. Meanwhile operating income stands at $2.1b (over 3x DCAI’s), though it’s down almost $500m from Q2’24.
Overall, Q2 is part of the weaker half of the year for client chip sales, so this is rarely an exceptional quarter for the group. While it’s not seeing any wild swings, the drop in both revenue and operating income are a sign that chip sales are down versus the year-ago quarter. Though with Intel noticeably cutting back on what details it reports about chip sales volumes and ASPs, we don’t have any formal figures on just what has happened with chip sales.
Based on this, we expect that Q2 was largely a repeat of Q1 for Intel. Continued ramping of Arrow Lake has shifted more of Intel’s high-end chip production over to the newer architecture. But it’s also shifted that production away from monolithic chips made entirely in Intel fabs and towards tiled chips that incorporate dies made at TSMC. The same is true with Lunar Lake, with Intel admitting that it’s a low margin part and they sell the on-chip memory at cost, rather than getting a profit from it. As Lunar Lake continues to ramp through 2025, Intel’s chip production costs seem to be continuing to go up, as they rely on an external fab and more expensive packaging techniques in order to integrate their silicon onto Intel chips.
Intel otherwise had very little to say about the CCG for the most recent quarter. With so much attention devoted to the foundry business and higher profitability DCAI products, the client group is continuing to move along at its regular pace.
The next big shift for CCG will presumably be the launch of Panther Lake late this year, which will be Intel’s first 18A-based product. Besides bringing some of that chip demand back inside Intel, Panther Lake will be critical in proving the viability and performance of the 18A node. Otherwise, desktop users will find themselves waiting a bit longer, as the first 18A desktop product, Nova Lake, isn’t set to launch until 2026. Intel reconfirmed that they expect a single Panther Lake SKU to launch this year, with the rest of the stack through the first half of 2026.
All Other (Altera, Mobileye, etc)
💵 Q2 Revenue $1053m, up 20% YoY from $881m
💰 Q2 Operating Income $69m, up from -$46m YoY
📈 Q2 Operating Margin 6.6%, up from -26.4% YoY
Intel’s final, catch-all business unit, ‘All Other’ which houses Mobileye (88% ownership*), Altera (moving from 100% now to 49%), IMS Nanofabrication (68%), and others, saw a good bump in both revenue and operating income. Revenue for the group has once again passed a billion dollars for the quarter, coming in at $1.053b, which is up 20% year-over-year. Operating income is a bit more mixed, however, as the $69m in income is a significant improvement over the year-ago quarter (where Intel lost $46m), but it’s a decline versus the previous quarter.
The small size and eclectic nature of the All Other unit means that it’s prone to more wild swings in revenue and income, and that’s essentially what we’re seeing here. For example, the FPGA business is seemingly still going through a period of post-pandemic digestion - other FPGA companies are starting to come out of this nadir several quarters later than anyone expected. This leaves the operating margin especially volatile, as evidenced here with swings between negative 5% and positive 14% within a year. Ultimately the companies in this business unit aren’t contributing a lot to Intel’s overall profitability right now – but perhaps more importantly, they’re not dragging it down, either.
More significantly, perhaps, is that the composition of the All Other group is set to change in Q3 and beyond. Intel’s previously announced sale of Altera to Silver Lake is expected to close in Q3, which will leave Intel as a minority shareholder of the FPGA company – a significantly reduced stake of what was once an internal Intel business unit. Though that sale should raise around $4b for Intel, ultimately shoring up Intel’s balance sheet at a time where the future of the company hangs in the balance.
Speaking of stock sales, Intel has also sold off about 7% of its remaining Mobileye shares, raising nearly $1b for the company. Intel remains Mobileye’s majority shareholder, so the company’s performance still has a significant impact on the All Other business unit’s performance. And with Mobileye seemingly doing well overall, Intel is hardly in a position to complain.
Outlook, Q2 2025
💵 Q3’25 Revenue, $13.1b, +/- $500m
📈 Q3’25 Gross Margins of 34.1% (GAAP) and 36.0% (non-GAAP)
Looking forward to the rest of Q3’25, Intel is currently forecasting a mixed quarter that will see total revenues decline slightly, but gross margins recover from this quarter’s particularly painful showing.
In terms of revenue, Intel is projecting between $12.6b and $13.6b in revenue for Q3. Going into what is normally the stronger half of Intel’s year, the midpoint of this estimate ($13.1b) has Intel seeing a slight amount of revenue growth QoQ, while revenue would be down around $200m on an annual basis. Intel’s guidance here is fairly shallow overall, but according to CFO David Zinsner, Intel is expecting the second half of the year to be weaker than usual. How much of this is the foundry side versus the product side remains unclear.
On the positive side, both GAAP and non-GAAP gross margins should recover from their recent lows. With a forecasted non-GAAP GM of 36%, the projected quarter is still not a great one for Intel, but it’s at least progress in the right direction. And compared to the especially dire performance of Q3’24, these figures would make Q3’25 look far, far better on a YoY basis.
Looking at Intel’s product plans for the next quarter, there’s also little change expected with Intel’s product lineups. Intel is now solidly in the middle of their product cycles for all of their major product lines – Lunar Lake, Arrow Lake, Battlemage, and Granite Rapids are all shipping, with many of these products set to continue ramping in volume in Q3. So while Intel’s chip mix continues to shift to newer products, there won’t be any major shake-ups for Q3 (at least among shipping products).
Intel’s next big development milestone remains the 18A process node and its associated GAAFET and Backside Power Delivery (BSPDN) technologies. As noted earlier, Intel has started 18A wafer production at its Arizona fabs, which puts them on time to supply tiles for the launch of Intel’s Panther Lake client part at the end of this year. A fuller launch will follow into 2026 for more Panther Lake parts, as well as the next generation Clearwater Forest Xeon processors in the first half of 2026.
Otherwise, the bulk of Intel’s forecasting within their Q2 earnings report is focused on their broader, long-term plans. Lip-Bu Tan previously set a goal of bringing Intel’s non-GAAP operating expenses for 2025 down to $17b, and according to the company they are on-track to hit that mark. Meanwhile gross capital expenditures are set to hit $18b for the year – another previously announced goal that Intel is on-track to hit – reflecting the reduced investments into Intel’s manufacturing operations.
Past that, nothing further has been announced about Intel’s spending targets for 2026. The company previously announced that they were aiming for just $16b in non-GAAP operating expenses in 2026, and with Intel on track to meet its 2025 targets, presumably they’re similarly on track to meet those 2026 targets as well.
Some Additional Commentary…
A small couple notes from Ian here.
On Wednesday, the White House unveiled its AI Action Plan - a way to approach the AI industry to benefit the country. It’s well worth a read, whatever you think of the administration. There are a couple of key points in there however, namely around ‘Next-Generation Manufacturing’ and ‘Restore American Semiconductor Manufacturing’. Both of these in context point to Intel and Intel alone as it relates to home-grown silicon development for leading edge AI compute, which should be a big win in terms of potential support, but it wasn’t mentioned in the call. Not only that, while AMD and NVIDIA were present at the event held in Washington DC, Intel was not. It’s something to flag with respect to LBT’s comments today on the call and the direction of Intel.
In terms of Intel at events on my radar, we have Hot Chips in late August at Stanford, where Intel is set to present about its E-core server product line as well as its new networking silicon. I’ll be there, say hi if you happen to be there too.
Also it’s worth noting that Intel postponed its Q3 2024 Intel Innovation event, stating it would be delaying it a year. Innovation is usually Intel’s chance to publicly align its vision with customers, partners, and press. At the time, Intel hasn’t confirmed this event is going ahead in 2025 either. As we get closer to the September/October timeframe, we should get more clarity. It would mark six months since LBT’s arrival, and it would be good to get that realignment.
I have more thoughts on today’s earnings and announcements, but I’ll put that in a separate piece tomorrow. It’s 1am here right now!
More Than Moore, as with other research and analyst firms, provides or has provided paid research, analysis, advising, or consulting to many high-tech companies in the industry, which may include advertising on the More Than Moore newsletter or TechTechPotato YouTube channel and related social media. The companies that fall under this banner include AMD, Applied Materials, Armari, ASM, Ayar Labs, Baidu, Bolt Graphics, Dialectica, Facebook, GLG, Guidepoint, IBM, Impala, Infineon, Intel, Kuehne+Nagel, Lattice Semi, Linode, MediaTek, NextSilicon, NordPass, NVIDIA, ProteanTecs, Qualcomm, Recogni, SiFive, SIG, SiTime, Supermicro, Synopsys, Tenstorrent, Third Bridge, TSMC, Untether AI, Ventana Micro.
Below the fold is for paid subscribers only.
It includes all the Lip-Bu Tan prepared remarks as well as the Analyst Q&A.
You could probably get these transcripts elsewhere, but that’s not here, is it? :)
