When did “we’re building chips for AI” stop being enough of a pitch? Because right now, every company with a soldering iron and a GitHub repo is claiming a seat at the agentic AI table. So when Barcelona-based Openchip says it’s targeting a 2028 launch on the back of that wave, the honest question isn’t whether that’s exciting — it’s whether there’s anything real underneath the announcement.
Let’s be fair. There is something here worth paying attention to. Openchip isn’t a slide deck with a logo. The company describes itself as a full-stack system-on-chip and software innovator focused on AI and high-performance computing. It’s building in-house designed, high-performance RISC-V based silicon — which, if you follow chip development at all, tells you they’re not just slapping someone else’s IP together and calling it a product. That’s a real technical direction.
What Openchip Actually Is
Openchip sits in a specific and genuinely interesting corner of the chip space. RISC-V is an open instruction set architecture that’s been gaining serious traction as an alternative to ARM and x86 — particularly for companies that want more control over their silicon without paying licensing fees to the big incumbents. Building a full-stack AI and HPC chip on RISC-V, with in-house design, is an ambitious move. It’s also a slow one. Chip development cycles are brutal, and 2028 as a target date isn’t aggressive — it’s realistic, maybe even conservative depending on where they are in the tape-out process.
The company showed up at MWC 2026 in Barcelona — including 4YFN and Talent Arena — which signals they’re actively positioning themselves within Europe’s tech ecosystem rather than operating in stealth. That’s a deliberate choice. You don’t go to MWC to hide.
The Agentic AI Angle
Here’s where things get strategically smart, or strategically convenient, depending on your level of cynicism. Agentic AI — systems that can plan, act, and operate with minimal human input — is genuinely driving new demands on hardware. These workloads are different from standard inference. They’re iterative, memory-intensive, and often need to run efficiently at the edge or in distributed environments rather than purely in massive data centers.
If Openchip is designing silicon specifically with those workloads in mind, that’s a real differentiator. The reports suggest their chips could affect AI computing supply chains and deployment costs globally, with a focus on lower-power operation. Lower power for agentic workloads is exactly the right problem to be solving. Running autonomous agents continuously is expensive — both in compute and energy — and anyone who can bring those costs down has a genuine market.
But — and this is a significant but — we don’t have detailed specs, benchmark numbers, or confirmed partnerships. What we have is a roadmap and a presence at trade shows. That’s not nothing, but it’s also not a shipping product.
Why Europe Needs This to Work
There’s a broader story here that goes beyond Openchip specifically. Europe has been trying to build a credible semiconductor and AI hardware industry for years, with mixed results. The EU Chips Act, various national initiatives, and a growing cluster of deep-tech startups are all part of that push. A Spanish company building full-stack AI silicon and targeting global deployment would be a meaningful data point in that story.
The supply chain angle matters too. Right now, the AI chip market is dominated by a very small number of players — primarily Nvidia, with AMD and Intel fighting for scraps, and Google’s TPUs serving internal needs. Any new entrant that can offer a credible alternative, especially one optimized for the specific demands of agentic workloads, changes the negotiating dynamic for every company building AI products.
My Honest Take
Openchip is doing the right things technically — RISC-V, full-stack ownership, HPC focus, agentic AI targeting. The 2028 timeline is plausible for a company at this stage. Showing up at MWC 2026 suggests they have something to show, even if the public details are still thin.
What I’d want to see before getting genuinely excited: actual silicon, even in early form. Power efficiency numbers. A named customer or design partner. Those are the signals that separate a well-funded roadmap from a product that will actually ship.
For now, Openchip earns a “watch closely” rating. The thesis is sound, the timing is right, and the European chip space badly needs a win. Whether this is that win gets answered somewhere between now and 2028 — and this reviewer will be checking back.
🕒 Published: