Intel 18A May Challenge TSMC, But 3-5 Million vs 17 Million Is the Lesson

The most important number in the Intel 18A story is not hiding under an electron microscope. It is sitting on the loading dock with a clipboard, asking how many chips can actually ship. Forbes frames Intel 18A as a possible TSMC challenger, and the architecture is genuinely interesting, but the more useful teardown starts with capacity. A transistor can win a benchmark slide; a foundry has to survive the production ramp without turning a customer roadmap into confetti.

The Spec Forbes Put Next to the Node Name

Forbes contributor John Werner says Intel 18A is positioning Intel as a potential challenger to TSMC, which Forbes describes as having long dominated the AI chip foundry market. The buried spec is the useful one: Forbes cites Intel current production volume at 3-5 million chips annually, compared with TSMC at 17 million. That is not a rounding error, it is the difference between one very impressive vault drill and a whole fleet of getaway vans. That volume gap is why process technology alone does not close a foundry race. A chip designer does not simply buy a transistor recipe; it buys predictable wafers, schedule confidence, ecosystem support, and the boring miracle of repeated execution. Forbes also argues that rising chip demand suggests room for two major vendors, which is the constructive read. Intel does not need TSMC to vanish from the field; it needs enough customers to believe Intel Foundry can become a dependable second path.

PowerVia Is the Fun Part, According to Forbes

Forbes says Intel 18A includes advanced gate all around transistors and PowerVia, a backside power delivery system, both aimed at improving performance and efficiency. Gate all around is the transistor version of putting a hand on every side of a wobbly ladder, better control when the geometry gets cranky. PowerVia is the heist movie bit: instead of making power and signals fight through the same crowded hallway, the design moves power delivery to the backside. That matters because modern chips are tiny cities with terrible traffic. Every logic block wants current now, every signal wants a clean path, and every bit of resistance or congestion becomes heat, delay, or both. Forbes notes that TSMC also uses gate all around in its N2 design and is developing a similar backside power solution. So the lesson is not that Intel has magic and TSMC has a paperclip; it is that both companies are chasing the same physical bottlenecks from different positions of manufacturing scale.

The Customer Test Is

the Real Teardown TechPowerUp, citing Reuters, reported that NVIDIA tested Intel Foundry 18A but did not proceed to mass production. That is exactly how serious foundry evaluation often looks from the outside: not a wedding, not a breakup, more like an engineer showing up with calipers, thermal paste, and trust issues. TechPowerUp also noted that many customers evaluate alternatives to TSMC before committing to more capacity. TSPA Semiconductor adds another useful industry breadcrumb, saying Broadcom, Nvidia, and Intel collaboration news had emerged, while its SemiVision team understood Intel 18A to have Tier 1 and Tier 2 customers, many of them CSPs. It also describes Broadcom as a Tier 2 customer and links that partnership to OpenAI Stargate, while noting OpenAI was also collaborating with TSMC on A16. The practical takeaway is simple: the AI chip world is not monogamous. Big customers probe options because supply assurance is now part of architecture.

What They Did Not Mention in

the Keynote Forbes points to news of Apple using Intel chips as a factor that significantly boosted Intel stock, a reminder that foundry confidence can move faster than fab reality. But the hard work is still wonderfully physical: repeatable yield, available capacity, packaging coordination, design enablement, and customer support that does not evaporate when the first mask spin gets weird. If thermal throttling is a betrayal, an unstable foundry ramp is the prequel where everyone ignored the blinking red light. For readers, the clean way to watch Intel 18A is to separate three questions. Does the transistor and backside power story hold up? Do external customers commit beyond evaluation? Can Intel narrow the 3-5 million versus 17 million production volume gap Forbes cited without losing the trust battle along the way? If those answers improve together, Intel 18A becomes more than a good process node. It becomes a credible pressure valve for an AI hardware market that clearly wants more places to build.