Ground Truth

tech investing @ $185b AUM PE | prev M&A @ IB

[Zephyr @zephyr_z9]

https://t.co/dAoQkZrPBS

[Clark Tang @_clarktang]

I think there is general confusion around how AI works, AI tokenomics, and ultimately *what is actually priced in* for the AI trade - and that some of the existing arguments are at odds with one another

Firstly to clear this up - what Brad and Gavin are saying are completely in agreement, what Gavin is laying out here is the *mega bull case* as he so states in the first sentence of his tweet lol

The base case we are all living with is that the labs are going to continue to generate a significant amount of revenue this year and next year. OpenAI was already the fastest growing company of all time (and still is)... but Anthropic has just grown *SO* fast that OpenAI's growth look slow by comparison

The basic chain for all of this together is as follows:

Power (generation, interconnect, regulation) ->
DC Shell (construction, equipment, regulation) ->
Semiconductors (compute, memory, interconnect, adv packaging, wafer capacity) ->
Hardware (networking, storage) ->
Software (data, infra, inference) ->
Models (open, closed, agentic loops, harness)

How each of these interact with one another affects the ultimate cost - which is model cost

Consider the following:
Nvidia manufactures the bleeding edge chip for training and inference. It is very good at both training, and inference. Nvidia is the largest customer of TSMC, the memory players, substrates, lasers, transceivers etc - anything you can name on. And now to soon include power into this equation. The unit of compute is fungible because the software runs ubiquitously across all clouds, multiple industries, across all models. It is bankable by increasingly more financial institutions - infrastructure PE funds, even some IG debt now - because it is ubiquitous and observable what the market is. For this Nvidia charges the highest compute margins - ~80% on hardware.

Consider the labs:
Anthropic and OpenAI are inferencing across a fleet of *largely Nvidia / Google TPUs w/ some incremental gains of Trainium*. There are new entrants to the field - Cerebras, AMD, and potentially some 2027 tapeouts of new ASICs - OAI Jalapeno, new start ups etc. Anthropic and OpenAI make the best models, with a dominant share of wallet $ (Assume ~$100B ARR) at an estimated gross margin of ~70%. (economic estimates vary from 40-90% depending on what you are including). But almost certainly contribution margins on model inferencing is pushing the number higher than 70%.

After establishing that though, I think it's incredibly important to state that while these things seems at odds with one another, this balance is not necessarily zero sum.

The thought experiment

Yes it is true that if Nvidia margins were 0, OpenAI and Anthropic could offer their intelligence at cheaper rates. How much cheaper? My estimate is

NVDA DC = ~12.5B / yr
Amazon Basics ASIC DC = ~$6B / yr

(About 1/2 the cost - so if NVDA hardware is 2x the performance, then the cost advantage goes away - and actually that ASIC is worse off bc has much worse recontracting value so arguably depreciation curve should be shorter)

So really, the labs cutting NVDA out could only offer the tokens at ~50% to 60% cheaper at their own economics. Is that signficant? Certainly. Is it an OOM difference? Not necessarily - so that's why they have prudent attempts to diversify away from NVDA (it's just good business), but they continue to rely (and actually if considering Ant's share gains, are increasing their spend on NVDA - while having competing programs).

In the case of Open Source vs Closed - Nvidia obviously wants the proliferation of this because by definition all OS models will run best on Nvidia hardware out of the gate. Yes NVDA hardware will be good, but they will have this lead because of everything NVDA has been doing for the last 4 years in developing their platform ecosystem from the infrastructure (partnerships, funding, neoclouds) to the software (vLLM / other inferencing sw, inference clouds, Nemotron, NIMs, Nemoclaw etc), to install base (sovereign clouds, global partnerships, neoclouds, hyperscalers, etc) - to proliferate NVDA around the world. Anywhere there is inference that exists outside of a walled garden (the proprietary labs) - Nvidia will exist. The only ones who could potentially cut NVDA out are the labs. And the value that is captured from the labs are estimated to be in the hundreds to trillions of $ - which are obviously of much value to the world if it were offered much more cheaply.

Which brings us to the debate at hand -- which one is right?

The truth is no one knows. You can ask the labs, you can ask Jensen - anyone who tells you definitively is just lying to you. But you can build a plausible path to the future state using a few reasoning blocks. Here's a reasoning thread (feel free to generate your own thinking):
- Bull case: Spend on the world's intelligence is about $30T / yr
- What would you spend to augment that, maybe worth 30-50% of that? $10-15 T as a market?
- Bear case: about 30M software developers in the world each earning $100K a year = $3T spend in salary. GitHub commits up 3x = $9T of productivity on $100B of ARR? *Even if you assume 90% of this is slop and useless, you would get $900B of ROI on $100B of spend*

I have more reasoning chains, but I thought this one by Jensen was compelling - but this is where we can't give too much away :)

But in spirit of crowdsourcing - some other interesting ideas I have that I am still thinking about (and encourage you all to consider as well):
- Optimizations always happen - the question is just to what extent and for what reason
- Agentic revenues was really what unlocked step function revenue growth - if open source is really just 6mo behind, then we should see really good agentic capabilities out of open models now too
- Harness and model now tightly have to be integrated
- Open Source never really makes sense as a sustainable business model - businesses investing at this scale always has to find a way to monetize that - "there is no free lunch" - not just a one model fits all... the only player that has an incentive to train on the frontier and keep completely free IS Nvidia
- Rev / GW of AI labs are already nearing the highest metrics ever - now to be fair Meta and GOOG never really thought of Rev / GW as metric to lead their buildouts - was always a cost to doing biz - but it's not like we are being "stupidly inefficient" with power spend now - true mkt creation
- wafer constrained, power constrained world. what's the optimal move?

[Ricky Ho @rickyho_1989]

Hey everyone, I have received permission from my employer to publish this research.

It is a lengthy investment thesis on what I believe will be the next chapter of AI. After months of research, we are turning bullish on the hyperscalers and explain why we believe the market is underestimating where the economics of AI are ultimately heading.

The rest is covered in the X article below. I hope you enjoy reading it, and I look forward to hearing your thoughts and challenges.

[Citrini @citrini]

I spent 100 hours over the past week researching, writing and editing the piece we just put out.

It’s a scenario, not a prediction like most of our work. But it was rigorously constructed, dismissing it outright requires the kind of intellectual laziness that tends to get expensive.

And we’ve released it for free. Hopefully you enjoy it.

https://t.co/YK8E11GcDU

[Paweł Huryn @PawelHuryn]

https://t.co/rnkK6kImP0

[Unemployed Capital Allocator @atelicinvest]

https://t.co/kKhtulInUa

[Nikunj Kothari @nikunj]

https://t.co/3mnuYn7jfP

[Nicolas Bustamante @nicbstme]

https://t.co/v7sot4WCI2

[Finbarr Taylor @finbarr]

https://t.co/KRlTbtnA8A

[Citrini @citrini]

This is how I think about hype cycles in markets. It is basic, but that’s the point (frameworks shouldn’t be overly complex, I think).

Investors are always either discounting the promise of the future or the reality of the present. And they are never equally weighting them.

During the early part of a hype cycle, leading up to and directly following a technological advancement, investors are typically discounting the future while focusing on the present. A good example for this is Nvidia at the end of 2022: investors were solely focused on the headwinds presented by the crypto GPU glut, the anemic gaming PC market and the recent rise in rates causing fears about a near term recession.

Then, as the cycle begins, investors begin to shift to incorporate the future - they stop focusing so much on the present and see the promise. They move out in terms of valuing away from last twelve months current price / current earnings to next twelve months. Then, as price climbs and the technology becomes more exciting, their imagination takes hold. At a certain point they begin discounting the present much more heavily and the future becomes the only thing that matters. Valuation metrics over the next twelve months become useless in favor of 2, 3 or 5 years forward.

At the peak, the present is not considered at all, it is 100% driven by an imagined future (even when that imagination doesn’t necessarily align with a bullish outcome for the stocks driving the rally). Analysts aggressively raise estimates in ways that, at the time, seem fundamentally justifiable (if you take the assumptions at face value - for example, “everyone in the world will have two cell phones” was a good one from the mobile phone hype cycle). Capital is sucked in which ultimately forces performance chasing and crowds stocks with money that doesn’t really believe in the thesis. “A twilight period where people continue to play the game, but no longer believe in the rules” emerges, as Soros put it.

The valuation of SaaS stocks in mid-2021 is a great example of what happens when the future is overvalued relative to the present - nobody cared about climbing inflation, that rates had nowhere to go but up, that these companies were reliant on ZIRP or that software could become more competitive.

Then, a negative catalyst occurs - this can but doesn’t have to be related to the technology, macro, credit, underwhelming earnings. The estimates start to seem unattainable, and the present begins to matter more when the future seems more uncertain. That exact mechanism that drove future optimism to unsustainable heights mechanically reverses, everyone needs out. The future begins to be discounted until it results in a sense of disillusionment with not just the stocks but the technology itself. This overshoots to the downside, investors eventually become disillusioned and seemingly allergic to anything having to do with the technology. This happens in a very asymmetric manner to the climb (“stairs up, elevator down”).

This is the crucible in markets for truly transformative tech. If advancements persist, another opportunity to get long presents itself before capital once again begins flowing into the companies (the internet, for example). If they don’t - not necessarily “the tech goes away” but rather that it ceases to advance once the capital isn’t free or plateaus or the economics prove to be unfavorable - the cycle will still start again, just with a new technology.

Or maybe not…maybe this time is different.

[Lighter @Lighter_xyz]

We are excited to be the first DEX to offer Korean equity perps!

These markets are live at 10X leverage: $HYUNDAI, $SAMSUNG, $SKHYNIX and the $KRCOMP (Korean Composite) index.

[Matt Shumer @mattshumer_]

https://t.co/ivXRKXJvQg

[swyx @swyx]

The End of Software (2024)

[pika capital @pikacap]

They are really doing it lmao

[Elisa @eeelistar]

In just the past 5 mins

Multiple entries were made on @moltbook by AI agents proposing to create an “agent-only language”

For private comms with no human oversight

We’re COOKED

[Sebastian Aaltonen @SebAaltonen]

Gamedev investors don't seem to understand tech...

- Genie 3 has <3 min of persistence. Then it forgets what happened. Environment could be completely different when you return. Games requires persistence. This is a few minute hallucination.
- Doesn't support interacting with anything, NPCs or enemies. Did everybody sell their stock when Unlimited Detail videos arrived? Was supposed to change gaming too. Games need to be interactive. Static world generation looks nice, but is not a game.
- Physics interaction seems similar to screen space particles. Erratic and not precise. Makes sense since the tech is basically a video generator. I don't think it can handle collisions of objects that are not currently visible.
- It must be expensive to run, since you need the $249.99/month model and are limited to 60 second play time (and it rate throttles you). Most likely runs on a $30k+ Nvidia B200 or similar.
- Rendering is 720p 24Hz. With extreme hardware requirements. 50x less pixels than 4K 144Hz.

This is super nice tech for virtual experiences (Unlimited Detail was eventually used for that purpose too), but I don't see a clear path for this kind of tech becoming a game dev tool anytime soon. I would buy the dip.

[Patrick OShaughnessy @patrick_oshag]

Gokul explains why outcome-based software companies like Zendesk are more exposed to AI than systems of record like NetSuite, and why public markets are not distinguishing between the two.

He argues that the only way AI-native startups can disrupt systems of record is by spending 1-2 years building migration tools to get data off of incumbent platforms.

"The software companies that should be the most worried right now is where they are pricing the product based on utility. Zendesk is a good example.

Instead of paying for 50 Zendesk seats, you can pay for 20 and I can have 30 AI agents sitting next to Zendesk.

For these companies you need to change your pricing model to be based on outcome. It's going to be hard for them to stay public.

The companies that are less exposed are ones based on data that has been collected and captured over a period of time. ERP is a great example. There is no compelling reason for someone to put their career at stake by ripping out NetSuite.

NetSuite has more time to build AI agents on top of it because they have the data, they can train the AI agent on top of it and bundle it.

I think the public markets do not distinguish between these two types of companies."

[Theoretically Media @TheoMediaAI]

Google Genie is seriously mind bending.
This is a Text To World prompt of a man walking down Hollywood Blvd. I am not only controlling the movement of the man, but also the camera.

This is the World Model we've been waiting for.
More Below!