TBPN's Run of Show: Chip Bottleneck vs. Energy Bottleneck
Claude Code starting to melt minds, the SaaSpocalypse has investors shook, Amazon posts eye-popping capex forecast
Happy Friday.
We’re live from YouTube and 𝕏.
The current thing in tech and business is people feeling the AGI with Claude Code.
Today’s lineup
SemiAnalysis President Doug O’Laughlin at 12:30 PM
Affirm Co-Founder and CEO Max Levchin at 1:00 PM
Matrix General Partner TJ Parker at 1:20 PM
Databricks SVP of Field Engineering Arsalan Tavakoli at 1:40 PM
Daily Op-Ed, by John Coogan
Chip Bottleneck vs. Energy Bottleneck
There’s an interesting tick-tock going back and forth in the AI supply chain on what is the key bottleneck to growth. Truthfully, I think only the AI labs themselves really feel the bottleneck directly. Most users just think “My chat app responds like normal, so things are fine.” But scaling is real, capex is going through the roof, and the economics are holding strong enough to propel further investment across the industry. So for 2026, I thought we were going to be really focused on energy, and we probably should be able to solve the next bottleneck. But right now, it feels like chips are more important to talk about.
Sam Altman put it this way yesterday on the show:
It [the bottleneck] goes back and forth. Right now, again, it’s chips. It’s different at different times. It may get solved on its own. Normal capitalism may solve it. But I think somehow deciding as a society that we are going to increase the wafer capacity of the world and we’re going to fund that and we’re going to get the whole supply chain and the talented people we need to make that happen would be a very good thing to do.
So why do we have a chip bottleneck? A lot of it comes down to consolidation. Power plants, data centers, cooling technology, there are a bunch of suppliers in each of these industries. You can parallelize them and steer resources from adjacent areas to focus on AI projects specifically.
To be clear, getting energy production on a dramatically different scaling curve will be an incredible challenge. The semiconductor industry has been riding an exponential for decades. Energy infrastructure (at least in America) most certainly has not.
Dylan Patel added some great context about how the energy industry differs from the semiconductor industry:
The semiconductor industry is used to doubling the amount of transistors made every year or two. Part of that is more advanced nodes, part of that is more capacity. Whereas the energy industry in America wasn’t built for that kind of growth. So initially, people weren’t creative. They were like, “Let’s build these combined-cycle gas plants.” But now we’ve realized — yes, there are three main manufacturers of turbines, and for a dual combined cycle you’ve got IGTs, but you’ve also got medium-speed reciprocating engines. Turns out Cummins can make about a million diesel engines a year, and those can generate electricity. If I don’t care about aesthetics and I put them in West Texas — easy.
Leading-edge fabs are completely different beasts. They cost tens of billions of dollars and take three, four, maybe five years to go from breaking ground to actually producing volume. Just look at TSMC Arizona. It broke ground in 2020 and still didn’t deliver volume in 2025.
There are also bottlenecks within the chip bottleneck. ASML is the only viable producer of EUV lithography machines, they ship around 50 per year, each one costs $350 million, and leading-edge fabs need dozens. So if you want to build a bunch more fabs, you need a bunch more tool makers, and ASML has its own supply chain, which isn’t super diversified. Vendors are highly specialized.
Even after you get the fab built, there’s at least a year of process engineering required to get to high-yield production. TSMC has decades of intellectual capital locked in the heads of engineers that can’t easily transfer or parallelize. This has made TSMC the real bottleneck. Hyperscalers are pushing capex numbers into the hundreds of billions, but the supply curve for leading-edge wafers is relatively inelastic. TSMC controls 90% of the advanced node market, with Samsung and Intel far behind.
This is why Ben Thompson is urging the tech companies who buy chips from TSMC to wake up:
The reality that hyperscalers and fabless chip companies need to wake up to, however, is that avoiding the risk of working with someone other than TSMC incurs new risks that are both harder to see and also much more substantial. Except again, we can see the harms already: foregone revenue today as demand outstrips supply. Today’s shortages, however, may prove to be peanuts: if AI has the potential these companies claim it does, future foregone revenue at the end of the decade is going to cost exponentially more — surely a lot more than whatever expense is necessary to make Samsung and/or Intel into viable competitors for TSMC.
Energy is still important, and that whole industry has an important job to do now to make sure the world is prepared for the terawatt-level demands of the 2030s, but there are bigger fish to fry this year.
Headlines
SemiAnalysis: Claude Code is the Inflection Point
WSJ: Long-Running AI Agents Are Here
WSJ: The Week Anthropic Tanked the Market and Pulled Ahead of Its Rivals
CNBC: Goldman Sachs is tapping Anthropic’s AI model to automate accounting, compliance roles
Reuters: Amazon sees 50% boost to capital spending this year, shares tumble
WSJ: White House Debuts Drug-Buying Site TrumpRx, With Roughly 40 Medications
WSJ Video: Republican Megadonor Ken Griffin on Trump’s Economy
CNBC: Novo Nordisk says it will take legal action after Hims & Hers reveals $49 copy of Wegovy pill
Bloomberg: Y Combinator Reverses Decision to Stop Investing in Canada Firms
WSJ: Behind Disney’s Search for a Lasting Successor to Bob Iger
Crowdfund Insider: In Run Up to Superbowl, Polymarket Hit with Class Action Lawsuit, Filed in SDNY
WSJ: A Michelangelo Drawing Just Sold for $27.2 Million—Over 13 Times Its Asking Price
Ramp releases list of top SaaS vendors
WSJ: The Utah Mom Who Got the World Hooked on ‘Dirty Soda’
The Supercar Blog: Tesla Roadster trademark reveals glimpse of its new design
WSJ: Miami’s Coconut Grove Has Become the Next Playground for Billionaires
Posts of the day
TBPN is brought to you by Ramp.
Ramp is the all-in-one finance automation platform that helps businesses save time and money with smarter corporate cards, spend management, and bill pay.
Special thanks to our sponsors: Shopify, Graphite, ElevenLabs, CrowdStrike, MongoDB, Fin, NYSE, AppLovin, Phantom, Gemini, Cognition, Labelbox, Public, Kalshi, Restream, Vanta, Console, Railway, Figma, Plaid, Okta, Linear, Turbopuffer, Lambda, Gusto, Vibe.co, Sentry, and Cisco.