'''
The Counterintuitive Truth About AI Chip Design
Everyone is obsessed with the latest AI models. GPT-5, Claude 4, whatever comes next. But nobody is talking about the brutal reality of the hardware that runs them. And that’s where the real winners and losers are being decided.
I’ve spent my career in the trenches of Silicon Valley. I’ve built and sold companies, and now I invest in the next generation of founders. I’ve seen firsthand how the right AI hardware can be a rocket ship, and the wrong hardware can be a lead balloon. I’m here to share the hard-won lessons and contrarian insights I wish I had when I started.
The Great GPU Lie
I once advised a startup that had raised a $50 million Series A. The founders were brilliant, with a groundbreaking new model for protein folding. Their first move? They spent $10 million on a massive cluster of NVIDIA H100s. Six months later, they had burned through half their cash, and their model was still not production-ready. They were so focused on having the "best" hardware that they never stopped to ask if it was the right hardware.
This is the great GPU lie. The belief that more GPUs, and only GPUs, are the answer to every AI problem. It’s a trap that I see so many startups fall into. They get caught up in the hype, the benchmarks, the endless race for more teraflops. But they forget that hardware is a means to an end, not the end itself.
Custom Silicon: The Unfair Advantage
In 2019, I was working with a company in the autonomous vehicle space. They were hitting a wall with off-the-shelf GPUs. The latency was too high, the power consumption was through the roof, and the cost was astronomical. I gave them some advice that sounded crazy at the time: build your own chip.
It was a huge risk. It took them two years and $30 million to get their first chip back from the fab. But it was the best decision they ever made. Their custom ASIC was 10x faster and 100x more power-efficient than anything they could buy off the shelf. It became their unfair advantage, their moat. It’s a huge part of why they are a leader in their field today.
Building custom silicon is not for the faint of heart. It’s a long, expensive, and unforgiving process. But if you have the right team and the right problem, it can be the most powerful weapon in your arsenal.
The Future is on the Edge
Everyone is still focused on the cloud. But the real action is on the edge. The next generation of AI applications will not be running in massive data centers. They will be running on your phone, in your car, in your home. And that requires a completely different approach to hardware.
Edge AI is all about efficiency. It’s about packing the most performance into the smallest power envelope. It’s about designing chips that can run for years on a single battery. This is where the real innovation is happening. Companies that are still stuck in a cloud-centric mindset are going to be left behind.
TPUs and Quantum: Don't Believe the Hype
I get asked about TPUs and quantum computing all the time. My answer is always the same: don’t believe the hype. For 99% of companies, they are a distraction. Google’s TPUs are impressive, but they are designed for Google’s scale and Google’s problems. They are not a magic bullet for your startup.
And quantum computing? It’s a fascinating area of research. But we are still decades away from having a quantum computer that can solve a real-world problem. It’s a science project, not a business tool. Focus on the problems you can solve today, with the hardware you can buy today.
The Only Thing That Matters
At the end of the day, the only thing that matters is finding the right tool for the job. Don’t get caught up in the hype. Don’t follow the herd. Do the hard work of understanding your problem, your constraints, and your goals. And then, and only then, should you start thinking about hardware.
The counterintuitive truth about AI chip design is that it’s not about the chips. It’s about the problems. The companies that understand this are the ones that will build the future. '''
Frequently Asked Questions
Do all experts agree with this view?
No, and that's fine. The best ideas in business are often contrarian. I share my perspective based on my experience and data, but I encourage you to seek out opposing viewpoints and form your own conclusions.
How has this view evolved over time?
My thinking on most topics has changed significantly over the years. Early in my career, I held many conventional views that experience proved wrong. I try to update my beliefs when the evidence changes.
What experience informs this perspective?
This perspective comes from over a decade of building companies in Silicon Valley, two successful exits (RemoteTeam to Gusto, MovieLaLa to Gfycat), and investing in 200+ startups including Anthropic, OpenAI, and Scale AI. I write about what I've lived.