AN-096 Company lists · list
Semiconductor Startups in the Bay Area
Semiconductor startups in Bay Area cities from the HardwareMap index: AI accelerators, photonic links, RISC-V IP. HQ, stage and disclosed raise.
by the HardwareMap editors3 min read
From the HardwareMap index, REV 26.09, refreshed 8 Oct 2026
Contents
Semiconductor startups in Bay Area cities number 16 in the HardwareMap index as of REV 26.09, from San Jose to Emeryville, all designing chips or chip-based systems rather than running fabs. This page lists each with HQ, stage, disclosed raise and status, sorted by disclosed capital. It is the regional slice of the AI chip startups list, which covers the field worldwide.
The table is live: it reads from the same records as each company's datasheet. To see every Bay Area chip company the index adds between refreshes, filter the live index.
Ranking rule: disclosed capital raised, largest first. That is a measure of investor conviction, not of chips shipped, so read it next to the product column on each datasheet. "Raised" means disclosed rounds only. Where a round is not public, the table says undisclosed and we do not estimate. Stage comes from the last disclosed round. Team size, where shown on the datasheet, comes from public profiles in September 2026 and moves faster than the other fields.
Where semiconductor startups in Bay Area cities sit
The list clusters on the south side of the bay. San Jose holds five entries, Santa Clara four, Mountain View three and Sunnyvale two. Palo Alto has PsiQuantum, and Emeryville, across the bay, has Oxide Computer Company. That is the same corridor the established chip companies built, and it is where foundry, EDA and packaging teams already work.
Founding years run from 2015 to 2022. SiFive, Ayar Labs and Cerebras Systems are the oldest. Etched, MatX and EnCharge AI are the youngest, all founded in 2022, all aimed at large language model workloads.
AI accelerators: eight of sixteen
Half the list builds compute for AI models. Each takes a different bet on where the bottleneck is.
- Wafer-scale. Cerebras builds processors from an entire silicon wafer and sells them as the CS-3 system (Cerebras).
- Hard-wired transformers. Etched builds Sohu, an inference chip fixed to the transformer architecture (Etched). MatX builds MatX One, an accelerator for large language models.
- Dataflow. SambaNova Systems sells reconfigurable dataflow systems, currently the SN50.
- Compute in memory. d-Matrix builds Corsair, an in-memory compute chip for inference. EnCharge AI takes the analog route with its EN100 accelerator.
- Open instruction set. Tenstorrent pairs RISC-V cores with AI processors, ships Blackhole, and publishes its software stacks as open source (Tenstorrent).
- Inference systems. Tensordyne sells Napier, an inference compute system, with automotive among its target markets.
Investor overlap is visible in the index records. Jane Street appears on both Etched and MatX. Spark Capital appears on MatX and on Enfabrica. Fidelity appears on Cerebras, Lightmatter and SiMa.ai.
Interconnect, IP and systems
The second group moves data rather than computing on it. Lightmatter builds Passage M1000, a photonic interconnect that links AI chips with light. Ayar Labs builds TeraPHY, an optical I/O chiplet that replaces copper links between chips. Enfabrica builds networking silicon and SuperNICs for AI data centers. As AI clusters grow, the link between chips becomes the constraint, and three Bay Area entries are built around that claim.
SiFive licenses commercial RISC-V processor cores to other chipmakers (SiFive), so its silicon ships inside other companies' products. Axiado builds hardware-anchored security silicon for AI data centers. Oxide sells rack-scale servers with hardware and software built together. SiMa.ai is the only edge entry: Modalix is a system-on-chip for robots and other physical AI.
PsiQuantum is the outlier. It builds photonic quantum computers, its system is called Omega, and it states that its chips come out of commercial fabs rather than custom lines. It sits in the silicon category because the product is a chip process.
What the Bay Area list does not include
The index catalogues startups with a product, a public source and an operating team. Large incumbents such as NVIDIA, AMD and Intel are not startups and are not listed. Design-service firms and pure EDA software companies are also outside the index.
Silicon startups elsewhere in California and the US sit in the silicon category: Syntiant in Irvine, Kneron in San Diego, Mythic in Austin and Positron AI in Reno among them. For the full Bay Area hardware picture, including robots, energy and aerospace, see hardware startups in the SF Bay Area. For how chip companies raise, see hardware startup funding.
Frequently asked questions
How many semiconductor startups are in the Bay Area?
The HardwareMap index lists 16 silicon companies headquartered in the SF Bay Area as of REV 26.09. Five are in San Jose, four in Santa Clara, three in Mountain View, two in Sunnyvale, one in Palo Alto and one in Emeryville. Wider counts run much higher because they include design-service firms and IP shops with no product of their own.
Which Bay Area chip startup has raised the most?
By disclosed capital in the index, Cerebras Systems leads, followed by Tenstorrent and SambaNova Systems. All three build AI compute. Disclosed funding measures investor conviction, not chip volume or revenue, and several companies on the list have rounds that are not public. The table on this page sorts by disclosed raise so the order is visible.
Do Bay Area semiconductor startups make their own chips?
None of the 16 runs its own leading-edge fab. They design chips and have them made by foundries, a model called fabless. PsiQuantum states that its photonic quantum chips are made in commercial chip fabs. Oxide Computer designs whole rack-scale servers rather than a single chip, and SiFive licenses processor cores to other chipmakers.
What kinds of chips do Bay Area startups build?
Mostly AI compute. Eight of the 16 build accelerators or inference systems for AI models. Three build interconnect: photonic links, optical I/O chiplets and networking silicon. The rest cover RISC-V processor IP, edge AI chips, security silicon, rack-scale servers and photonic quantum computing. Edge and automotive chips are a small share of the Bay Area list.
Is Cerebras a startup?
The index keeps Cerebras Systems in the silicon category because it was founded in 2015 and still sells one product family, the wafer-scale CS-3 system. It is far larger than most entries, with more than a thousand staff per the index. Its status field is tracked monthly, and its datasheet page shows the current record.
HardwareMap catalogues every Bay Area chip startup that meets the bar. Open any company's datasheet from the table for its product, team size and investors.
The index
All 219 startups, filterable by category, hub, stage and funding.