Intel has made a move. After two years of losses and a return to profitability supported by external factors (NVIDIA, SoftBank, and public aid in the US), the company has confirmed that it will prioritize the production of AI infrastructure chips over consumer CPUs and GPUs for PCs and laptops. This message was delivered during the Q3 2025 earnings presentation, signed by the new CEO, Lip-Bu Tan, and the CFO, David Zinsner: the manufacturing capacity that is currently limited will be sharply focused on AI servers, where margins are higher and demand continues to grow unabated.
“We expect CCG (consumer chips) to decline slightly, while DCAI (servers) will grow strongly because we will prioritize server capacity over entry-level processors,” explained Tan.
“Panther Lake will be a pretty expensive product at the start, and during the first half of the year, we will need to give more attention to Lunar Lake,” added Zinsner.
This decision significantly impacts consumer product plans and the schedule for Intel’s next-gen nodes (18A and 14A), painting a picture of 2025–2026 where AI leads and PC development takes a backseat.
Scarcity of silicon and margins: why PC production is on hold
Intel lacks sufficient capacity to meet total demand. In this context, it is focusing on what promises the highest profitability today: server CPUs and GPUs for training and inference AI clusters. This move aligns with market trends: Gartner already predicts that optimized AI IaaS will be the new engine of cloud services and that AI inference spending will surpass training from 2026. Operationally, each wafer diverted to DCAI yields better margins than one allocated to CCG.
For end users, this translates to less volume of consumer CPUs and GPUs in 2025, more limited launches, and, likely, higher prices where supply is constrained. In other words: Intel is consolidating its position in AI infrastructure and buying time with the client, waiting for its next-gen nodes to reach performance levels that improve costs.
18A: enough to supply, insufficient for margins; 14A exceeds expectations
The Intel 18A node is the core of this transition and, at the same time, its Achilles’ heel. According to Zinsner, the performance of 18A is “adequate to meet supply”, but “not where we need it” to achieve acceptable margins. Intel itself recognizes that 18A won’t be an immediate financial driver, and new investments will depend on “demand commitments from external clients”. In plain terms: capacity expansion will not occur without clear agreements, and currently, interest is limited.
Nevertheless, 18A will form the foundation for at least three generations of chips — client and server — and will serve as a common platform into 2027. The most tangible commercial detail: Panther Lake, the first 18A product for the client segment, will arrive in 2025 with only one version, and the rest of the product range will scale in 2026. Lunar Lake thus acts as a “bridge” during the first half of 2025.
The surprise comes from 14A: Zinsner assures that the node’s launch is better than 18A’s at the same maturity point, in terms of performance and yields. If confirmed, 14A could ease pressure in the medium term and accelerate the deployment of future families, though the company has not yet disclosed which specific products will migrate first to 14A.
Implications for product timeline: Panther Lake is “short,” and Lunar Lake comes to the rescue
Intel’s 2025 PC plans will be contained. Panther Lake will debut in 18A with a limited launch (one version) and a high initial cost, relegating it to select segments. To fill the gap, Intel will push Lunar Lake — focusing on efficiency, NPU, and autonomy — during the first half. If 18A improves performance towards the end of 2026 as Zinsner suggests, Panther Lake’s product lineup will expand, and adoption of 18A will begin to normalize.
In consumer GPUs, the message is similar: without concrete family details, the capacity remaining after meeting DCAI will determine availability and pricing in gaming and creative applications.
AI or PC: what benefits for the end user?
In the short term, gamers and laptop enthusiasts will see fewer major launches, more continuity, and a pricing pressure that was already anticipated with the series Core 12/13/14. In the medium term, if 18A and 14A mature and capacity increases, the PC could benefit from the cost decline and the innovations driven by AI (enhanced NPU, bandwidth, and efficiency).
The underlying question —“Is everything AI, and does the desktop matter?”— doesn’t have a binary answer. The growth of AI in infrastructure is indisputable, and its margin explains the priority. But the PC remains a volume channel with brand value and ecosystem. The key will be how quickly Intel can make 18A (and later 14A) deliver economic performance that allows re-acceleration without sacrificing margins.
AMD and NVIDIA in the rearview mirror
While Intel reorders its priorities, AMD takes advantage with its offensive in AI servers (Instinct family) and consumer CPUs and APUs, gaining traction in both data centers and PCs. Meanwhile, NVIDIA continues to dominate AI acceleration and software, with its ecosystem (CUDA, NVLink/NVSwitch) shaping the market pace. Intel’s margin for maneuver lies in executing its mix properly: serving AI today (revenue and contracts), stabilizing 18A margins, accelerating 14A performance, and returning to the customer with competitive offerings that incorporate NPU and efficiency without penalizing prices.
What to watch in the next 12–18 months
- 18A performance: if improvements happen by the end of 2026, customer spending could reignite in 2026–2027.
- 14A cadence: if it maintains a strong start, it could bring products forward or improve costs sooner than expected.
- External contracts (IFS): committed demand enabling capex to expand capacity without impacting margins.
- PC portfolio: real scope of Panther Lake and traction of Lunar Lake in ultraportable AI-PCs (autonomy, NPU performance, price).
- Average retail price: if the mix falls short, we will see higher PVP and irregular availability in certain ranges.
Conclusion: capacity management, not abandoning the PC
Intel is not “abandoning” the PC; it is managing a shortage by prioritizing where margins are today highest and most urgent. The success of this strategy will be measured in two trends: AI growth and contracts, and 18A/14A maturation that will eventually restore profitability to consumer products. If both converge, the PC can regain momentum with more competitive silicon; if not, AMD will continue to gain ground, and the user will feel that “everything is AI”… but with a desktop that’s becoming increasingly costly.
Frequently Asked Questions
What about Panther Lake and Lunar Lake?
Panther Lake (18A) will arrive in limited form in 2025 (one version only), with gradual deployment in 2026 as the node improves. Lunar Lake will be the mainstay in the first half of 2025.
Why prioritize AI servers now?
Because of margins and demand. Capacity is limited, and AI offers a better return than the consumer. Plus, the inference market is growing rapidly.
Is 18A “bad”?
Not exactly: currently, it satisfies the supply needs, but still doesn’t reach the yield levels that would enable attractive margins. Intel expects improvements by late 2026 and subsequent stabilization. In contrast, 14A shows a better start.
Will prices in PC and gaming go up?
It’s probable we’ll see greater price pressure and less volume in 2025, especially in segments where capacity is prioritized for DCAI. Normalization will depend on yields and contracts that enable more production.