Elliot Turner

Dad. RGA Investment Advisors. Soundview Equity Partners. GARP investor. Mets, Jets, Isles fan. Compulsive reader. In awe of nature. Co-host @twiii_podcast.

[Elliot Turner @ElliotTurn]

Incredible new episode with Doug O'Laughlin of @fabknowledge and @SemiAnalysis_!!

On how the semi industry is leading AI forward, the winners of the AI race, an intriguing investment opp in the newest bottleneck for the industry & the deep recession in the non-AI semi world.

[This Week in Intelligent Investing @twiii_podcast]

.@ElliotTurn and @mihaljevic welcome Doug O'Laughlin of SemiAnalysis and Fabricated Knowledge. Doug shares insights into semiconductors, how the industry leads AI forward, AI winners, opportunity in an industry bottleneck, and the non-AI semi recession.
https://t.co/jtmhlDynqj

[Elliot Turner @ElliotTurn]

Narrator: he was busier than we knew.

[Elliot Turner @ElliotTurn]

History rhymes: the COVID liquidity bubble popped when Archegos got Hwanged.

Four years later, $NVDA is busy Huanging the AI bubble in an entirely different asset class.

[Elliot Turner @ElliotTurn]

History rhymes: the COVID liquidity bubble popped when Archegos got Hwanged.

Four years later, $NVDA is busy Huanging the AI bubble in an entirely different asset class.

[The Inflection Economy @inflectionecon]

We’re out here in Chicago for Latticework 2026 by @manualofideas @peter_mantas chatting with @ElliotTurn about biopharma. The golden age of biotech is upon us….the industry is looking very different today than it did 10 years ago.

[Elliot Turner @ElliotTurn]

To be clear, I don't think AI per se is a bubble, but do think some bottleneck sticks are. I actually am not convinced NVDA is one of those.

I just thought the rhyme was funny in the karmic sense.

[Elliot Turner @ElliotTurn]

History rhymes: the COVID liquidity bubble popped when Archegos got Hwanged.

Four years later, $NVDA is busy Huanging the AI bubble in an entirely different asset class.

[Elliot Turner @ElliotTurn]

@David_Doran7 That's the short, focused list!

[Elliot Turner @ElliotTurn]

History rhymes: the COVID liquidity bubble popped when Archegos got Hwanged.

Four years later, $NVDA is busy Huanging the AI bubble in an entirely different asset class.

[Elliot Turner @ElliotTurn]

I couldn't get the thought out of my head so I had to map it out.

This chart shows the COVID winners from April 2020-Dec 31 2021 plotted against the AI Bottleneck 40 from April 2025-today.

Look at the overlay and look at the correlations, it's jaw-dropping!

[Elliot Turner @ElliotTurn]

@JH_Grandgerard To be clear, I don't think AI per se is a bubble, but do think some bottleneck sticks are. I actually am not convinced NVDA is one of those. I just thought the rhyme was funny in the karmic sense.

[Elliot Turner @ElliotTurn]

@WalmartHwang C'mon Bill it was right there for you!

[Elliot Turner @ElliotTurn]

History rhymes: the COVID liquidity bubble popped when Archegos got Hwanged.

Four years later, $NVDA is busy Huanging the AI bubble in an entirely different asset class.

[Elliot Turner @ElliotTurn]

@hareng_rouge @SkeleCap Or the steel for the tracks.

When steam goes electric, there's big opportunity though!

But applications > infrastructure.

[Elliot Turner @ElliotTurn]

Part of why I've been thinking about this is our belief that at some point AI needs a handoff from the "infrastructure layer" to the "application layer."

The value stops accruing to the component manufacturers as new tech powers new, previously unfathomable businesses.

[Elliot Turner @ElliotTurn]

I couldn't get the thought out of my head so I had to map it out.

This chart shows the COVID winners from April 2020-Dec 31 2021 plotted against the AI Bottleneck 40 from April 2025-today.

Look at the overlay and look at the correlations, it's jaw-dropping!

[Elliot Turner @ElliotTurn]

@HJB2_0 Lol he did kick his ass! Almost got the Gordie Howe hat trick, can't believe I missed the game.

[Elliot Turner @ElliotTurn]

Really excited for this with @peter_mantas at @manualofideas!

[Peter Mantas @peter_mantas]

I will be in Chicago next Wednesday (w/ @inflectionecon) to have a fireside chat with @ElliotTurn on the opportunity at present in biopharma and life science tools for generalist investors.

Will be a fun chat so don't miss out!

[Peter Mantas @peter_mantas]

I will be in Chicago next Wednesday (w/ @inflectionecon) to have a fireside chat with @ElliotTurn on the opportunity at present in biopharma and life science tools for generalist investors.

Will be a fun chat so don't miss out!

[Borlaug @Borlaug_]

Have been thinking about this as well. We've all had a front-row seat to the discussions and debates around the buildout, but once you get to use cases, the debate is fixated on coding and personal agents and drops off thereafter. Not as much around where broader enterprise adoption is. Found that surprising, especially because you need real enterprise adoption for any of this stuff to matter 5 years from now.

I picked up coverage of software, compute, and, power earlier this year and it was pretty jarring how far apart the software company debates were from and compute/power/semis/labs discourse.

If you follow the traditional SaaS businesses, especially the big system-of-record guys, it's clear that for most, real adoption and usage of AI products sold by the SaaS space is a late 2027 event at the earliest. $NOW and $TEAM are quickly realizing that they need to deploy engineers to customer sites like $PLTR. Both are telling investors to expect little to no revenue upside from customers using enough tokens to generate an overage beyond contracts. The "AI benefits" are contract expansions/upselling to plans that include it, an effort that companies seemingly believe must precede driving usage. Messy/noisy.

Then, if one looks at the infra SaaS players ($SNOW etc.), most of them aren't "benefitting from AI" the way you're thinking - they've just found new customers that spend at lot of money with them today. On the enterprise side, what you're seeing is not a big wave of AI usage demand, it's a big acceleration in cloud migrations that enterprises believe make them more "AI-ready". Migrations are often multi-year in nature. Cyber has elements of this as well.

Taken together, the logical reaction is "this means air pocket in 2027", but it's not like labs/the current apps aren't still growing rapidly, it's not 100% clear we're out of the woods on scarcity/buildout just yet, and some companies are farther along with their AI product launches than others.

That said, "air pocket" sounds scarier than it is. Think that's mostly because it implies deceleration, and it today like 80%+ of tech companies across both software + semis have a decelerating topline (most from growth rates that are attractive) and markets are punishing that right now.

In SW, it seems to be creating a lot of GARP-y profiles that are attractive in cases where one believes that terminal value is not going to 0. Patience while holding something cheap + durably growthy for a year gets rewarded in 2028+ with the benefits of acceleration etc.

[Elliot Turner @ElliotTurn]

@tierrapartners @hareng_rouge Yeah agreed that software and service have a role to play. Amazon has been my favorite of a company positioned to win the application layer, both the retail side w/ automation/robotics and AWS w/ suite of services built inside and around Bedrock.

[Elliot Turner @ElliotTurn]

@hareng_rouge But yeah, the applications are the future and that's the way it's supposed to be!

[Elliot Turner @ElliotTurn]

@hareng_rouge I do think there will be some infrastructure winners going forward who can take share but baby out with bathwater for now.

[Elliot Turner @ElliotTurn]

My thinking so far is that peak DC investment is tricky for a lot of bottleneck companies & the winners will be those who can provably take share w/in the DC rather than those who ride the wave. It's much tricker & the absence is painful even for the future winners.

[Elliot Turner @ElliotTurn]

@Borlaug_ @denmarkcap Cyclicals be cyclical even in a secular wave, yes!

[Elliot Turner @ElliotTurn]

I'm not sure that's all that different. The hyperscalers and frontier labs are not in my AI index here. I think those building recurring revenue streams are incomparable with those who supply to the buildout. This could be a very advantageous moment in time for $GOOG $AMZN $META

[Brandon Ladoff @denmarkcap]

one potential difference here is that, outside of coding, the underlying models / products could still be early in their utility and usage.