In partnership with

TAKEAWAYS FROM THE SILICON VALLEY AI STUDY TOUR

Last week, I delivered the opening keynote for the Silicon Valley AI Study Tour, a week-long immersion program organized by LinkSV and The Banking Academy, bringing 35 senior executives from 23 global banks into the heart of Silicon Valley to see firsthand what's actually happening in AI.
Their question was simple: what's real and what's hype? The answer wasn't what most of them expected.

Silicon Valley Isn't a Place. It's an Intersection.

Silicon Valley's real advantage is speed of information. This is where founders, operators, investors, and infrastructure builders physically overlap and share what they're learning in real time. That density of knowledge transfer doesn't exist anywhere else at the same pace.

The framing message: "The rate of learning must match the rate of change. Right now, this is the only place where that's consistently true."

For international financial leaders evaluating where to place AI bets, the takeaway was simple. You don't have to build here. But you need to be plugged in here.

Three Layers Moving at Three Speeds

I broke the AI market into three layers, each on a different clock:

• Software and applications - changing in days

• Models and tooling - changing in months

• Infrastructure - changing in years
  

Enterprise procurement, capital allocation, and regulatory frameworks all move on infrastructure timelines. But the technology they're evaluating shifts at software speed. For banks and institutional investors, the diligence cycle is often slower than the market itself.

Infrastructure Is the Bottleneck

The biggest constraints on AI scale right now are all at the infrastructure layer. GPU shortages. Energy shortages. Storage and memory limitations. And the sheer capital required to compete.

A viable neocloud needs roughly a billion dollars in GPUs and infrastructure. Right now we're advising on $100M+ raises that need to close within 90 days and billion-dollar targets over the next year just to secure enough data center capacity.

The training wave is already straining capacity. The inferencing wave, which will be bigger, hasn't fully arrived yet. Enterprise demand for private GPU clouds is slower than expected. But when it shows up, the gap gets wider.

The Iron Man Suit

I also walked through how we run IgniteGTM internally. 14 people. 10 active AI agents. I call the operating model the "Iron Man suit."

The human operator wears the AI. We iterate workflows multiple times a day. Humans stay in control. The agents handle execution, research, drafting, coordination. But every output gets reviewed, revised, and redirected by a person.

We've replaced our conventional CRM with agentic workflows. Not by cutting people, but by making each person way more capable.

What the Bankers Took Home

The delegation left with a practical filter for evaluating AI opportunities:

• Top-down projects driven by executive mandate or hype tend to stall

• Bottom-up projects driven by real experimentation tend to stick

I also shared a note on hardware strategy: don't chase the newest GPU generation in emerging markets. Older but capable hardware is often more practical and deployable.

On trust especially around Chinese models and open source, enterprises will pay a premium for vendors they trust. Price and performance matter, but trust is the deciding factor in regulated industries.

AI is the most important discovery since fire. But the path to capturing that value runs through infrastructure, trust, and the willingness to learn faster than the market moves.

Bay Area Startups Collectively Secured $2.1B+ in August Week 1

August started out slow, ending week one with just $2.1B in startup funding. The week's deals included fundings for AI infrastructure startups Buzz Solutions, Lumilens and Multibeam Systems.

SaaSpocalypse exits: There was one notable (but not for a good reason) acquisition this week – Bending Spoons' acquisition of Airtable for $1.28B. Their previous valuation was $11B in their 2021 Series F funding, $5.5B in the round before that and total funding of $1.36B. Not a win for the investors.

For startups raising capital: Drowning in 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, and the companies, investors and executives impacting the startup ecosystem.

Early Stage:

• Aurelius Systems closed a $40M Series A, an autonomous, directed-energy technology company building the defense ecosystem of the future.

• Sapiom closed a $35M Series A, provide the essential infrastructure (KYA, wallets, spending and usage controls, and multi-rail payments) to unlock secure, programmatic commerce.

• Naive closed a $28.5M Series A, provides the full stack your agent needs to act in the real world — compute, payments, identity, incorporation, 10,000+ tools — in one config file.

• Buzz Solutions closed a $20M Series A, an AI software company helping utilities modernize the electric grid through intelligent visual infrastructure management.

• Avatar Robotics Co closed a $6.5M Seed, building the unlimited workforce for industrial labor.

Growth Stage:

• Lumilens Inc. closed a $700M Series C, creates photonics products designed for high-performance computing.

• Mariana Minerals closed a $310M Series B, a software-first, vertically integrated minerals company focused on the minerals critical to modern energy, AI, and defense technologies.

• Multibeam Systems closed a $205M Series Unknown, helping accelerate semiconductor chip innovation with the first Multicolumn E-Beam Lithography system built for volume production.

• Expedition Therapeutics closed a $115M Series B, a biotechnology company developing novel therapies for serious inflammatory and respiratory diseases.

• Windborne Systems closed a $37M Series B, designs, manufactures, and operates the only constellation of smart, long-duration sensing balloons that power AI global weather forecasts.

Helios

Inference is becoming a much bigger infrastructure problem as AI moves into production. Every user request, agent action, generated image, and reasoning step consumes compute. At scale, teams have to think about throughput, latency, GPU availability, and what every inference request actually costs.

Companies are looking for a simple outcome: run models quickly and reliably without letting inference economics get out of control.

That is where Helios fits.

Who they are
Helios is an AI infrastructure provider offering GPU cloud, colocation, and serverless inference built around high density compute infrastructure.

What they deliver
Helios gives teams access to open models including Kimi, DeepSeek, Whisper, and Flux through serverless inference with per millisecond billing. That inference platform sits alongside dedicated GPU infrastructure, including Blackwell generation systems designed for large scale reasoning and production inference.

Who they serve
AI companies, developers, and infrastructure teams that need production inference without building and operating the underlying GPU infrastructure themselves.

As token volumes grow and inference becomes a larger share of AI compute demand, the ability to deliver that compute efficiently will have a direct impact on what AI products can economically scale.

Explore Helios to see how they are approaching production AI inference.

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

Write docs 4x faster. Without hating every second.

Nobody became a developer to write documentation. But the docs still need to get written — PRDs, README updates, architecture decisions, onboarding guides.

Wispr Flow lets you talk through it instead. Speak naturally about what the code does, how it works, and why you built it that way. Flow formats everything into clean, professional text you can paste into Notion, Confluence, or GitHub.

Used by engineering teams at OpenAI, Vercel, and Clay. 89% of messages sent with zero edits. Works system-wide on Mac, Windows, and iPhone.

Try Wispr Flow free