In partnership with

AMD just laid out its roadmap for the next era of AI compute and if you build, operate, or buy AI infrastructure, here's what matters.

1. The data center is the new unit of design

Dr. Lisa Su made it clear: the days of optimizing individual accelerators in isolation are over. Power, cooling, memory, and networking are now first-class design constraints. If your rack architecture doesn't account for all of them together, you're leaving performance on the table.

2. Helios is AMD's answer to rack-scale AI

AMD introduced Helios, a modular platform that integrates Instinct accelerators, EPYC CPUs, and Pensando networking at the rack level. The pitch: better performance per watt and simpler deployment across large clusters. For infrastructure operators, this is AMD signaling it wants to compete on full-stack system design, not just chips.

3. The accelerator market is heading past $200B

They project the data center AI accelerator market will grow at 50%+ CAGR. Demand is coming from more than hyperscalers now, enterprises, sovereign AI programs, research institutions, and emerging neoclouds are all driving spend. The buyer base is fragmenting, which means the infrastructure that serves them has to be more flexible.

4. The hardware roadmap extends through 2028

MI500 in 2027. MI600 in 2028. Both built on HBM4 memory architectures for larger training and reasoning workloads. For anyone planning data center capacity, power budgets, or capital deployment this is the planning horizon AMD is giving you.

5. Software openness is the competitive wedge

AMD is leaning hard into an open ecosystem play. In a market where CUDA lock-in is the default, AMD is betting that flexibility and portability will matter more as the buyer base diversifies beyond the hyperscaler top 5.

AI infrastructure is now a systems game; power, cooling, memory, networking, and software all have to move together. they’re positioning itself as the platform for operators who need choice, efficiency, and a roadmap they can plan around. Whether you're a hyperscaler, a neocloud, or an enterprise building out AI capacity, this is the competitive landscape shifting in real time.

We’re proud to be a community partner of the AI Summit at Stanford a three-day residential gathering for AI founders, investors, technologists, and researchers.

The summit runs July 30 to August 1 on Stanford’s campus.

Members of the IgniteGTM community receive 50% off tickets with code IGNITEGTM. Space is limited.

Bay Area Startups Collectively Secured $7.6B+ in July MTD

Funding activity stayed steady this week, but the median size of the deals went down from $24M to $11.5M. There were only five megadeals and the round size dropped from there, driven by a flurry of seed rounds that made up almost 50% of this week's fundings.

Exits: Out of ten acquisitions this week, only one was notable - Tempus acquiring Personalis for $1.5B. Scribe Therapeutics made their public debut in the first gene-editing IPO in more than two years, raising $129M after upsizing shares to 8.5 million and pricing at $15, the top of their range. The stock (SCTX) opened on their first day of trading at $25 and closed at $21.65.

For startups raising capital: Don't waste time on lists, find investors in companies like yours with LinkSV. Stay on top of who's raising, who's closing and who's investing with the Pulse of the Valley weekday newsletter. Founders get the newsletter, database and alerts for just $7/month ($50 value). Check it out and sign up here.

Follow LinkSV on LinkedIn to stay on top of SV funding intelligence, new tech and the companies, investors and executives impacting the startup ecosystem.

Early Stage:

• Natural closed a $30M Series A, building the financial rails for an agentic future where autonomous agents initiate, authorize, and execute payments.

• Elio closed a $21M Series A, their sensors let AI direct what to capture in real time.

• SkyPilot closed a $20M Seed, AI Compute Platform for frontier AI teams.

• Runta closed a $20M Seed, the execution layer for AI agents.

• Infinity.inc closed a $15M Seed, building the software layer that makes any chip competitive for AI inference.

Growth Stage:

• Meshy AI closed a $400M Series B, an artificial intelligence platform that enables users to generate and edit 3D models from text descriptions, images, and conversational prompts.

• Sila Nanotechnologies closed a $300M PE round, a next generation battery materials company dedicated to accelerating energy transformation.

• Etched closed a $300M Series C, builds frontier inference clusters designed to make AI inference dramatically faster, cheaper, and more abundant.

• Candid Health closed a $120M Series D, on a mission to rebuild the financial infrastructure of American healthcare

• Glow Security closed a $100M Series B, the Endpoint AI company with an AI-powered operating model built on prevention, not reaction.

  

AI inference is becoming more specialized. As models grow larger and AI agents generate more tokens, infrastructure teams are looking beyond one-size-fits-all architectures. The goal isn't simply adding more accelerators. It's matching the right hardware to the right workload.

That's the idea behind the new partnership between AMD and Cerebras.

AMD and Cerebras have announced a joint inference solution that combines AMD's Helios rack-scale AI infrastructure with the Cerebras Wafer-Scale Engine. The architecture pairs high-throughput processing with ultra-low-latency inference, allowing different parts of the workload to run on the hardware best suited for the job.

Rather than relying on a single platform for every stage of inference, operators are increasingly designing specialized, composable systems that optimize for performance, efficiency, and scale.

Learn more about how AMD and Cerebras are approaching the next generation of AI inference infrastructure.

Your Feedback Matters!

Your feedback is crucial in helping us refine our content and maintain the newsletter's value for you and your fellow readers. We welcome your suggestions on how we can improve our offering. [email protected] 

Logan Lemery
Head of Content // Team Ignite

Same AI software. Wildly different results.

Every company in this dataset bought the same AI capabilities. The difference in results came down to one thing: whether someone inside CX owned it.

One beauty retailer made 202 workflow updates in 30 days — refining as policies changed and new questions came in. Companies without a named owner saw performance stall or decline.

Read the data on what separates AI deployments that work from the ones that stall, and the four questions worth asking before your next AI investment.

See the data