Fernando Vidal

Cofounder and Chief Data Scientist at @3F_Research. Mainly trying to post cool charts. PM $FCTE $RAA

[Fernando Vidal @fernavid]

@TheStalwart Harnesses are useless if the models aren't good enough to not go off the rails as errors accumulate. Of course, you need the harnesses to actually allow people to consume huge amounts of tokens, but the models getting better are the fundamental driver of token usage exploding.

[Fernando Vidal @fernavid]

It's just a small sample of the total AI ecosystem, but OpenRouter AI token consumption growing almost 40% in the single month of May is pretty incredible.

[3Fourteen Research @3F_Research]

New report out to 3FR clients discussing:
-How the semi trade has influenced broad market structure,
-Loosening domestic GPU availability data BUT tightening foreign availability, and
-Why index price, internals, and valuations fall short of bubble characteristics.

[Gubb @gubbmintcheese]

Nice chart showing the wild, multi-decade disconnect between wages and house prices in Canada. Chart made in 1 min by Caliban of @3F_Research which is the amazing research firm run by @WarrenPies and @fernavid . Highly recommend you try the 7 day trial.

[Warren Pies @WarrenPies]

18 months later, the S&P 500 is:
-Above 7,000 ✅
-Still not overvalued ✅

More people agree w/ our valuation thesis now, but I still wouldn't classify it as consensus...

[Warren Pies @WarrenPies]

Joined @michaelsantoli to discuss market outlook and how the war has transformed the oil market.

https://t.co/8gpjfNwAmt

[Warren Pies @WarrenPies]

Talking oil and the need for an imminent off-ramp w/ @michaelsantoli this afternoon on @CNBCOvertime

Thank you for the conversation Mike!

https://t.co/XDuqKXtPIg

[Fernando Vidal @fernavid]

anthropic.APIStatusError: {'type': 'error', 'error': {'details': None, 'type': 'overloaded_error', 'message': 'Overloaded'}

$25 per million tokens and the API is overloaded from too much demand...

[Warren Pies @WarrenPies]

Went to the NYSE this afternoon to talk markets w/ @ScottWapnerCNBC @CNBCClosingBell

Oil still driving everything...next week is crucial.

https://t.co/xPpMU6m0bS

[Fernando Vidal @fernavid]

@stevehou @MetacriticCap Yeah agree these lab valuations take as a given that AI revenues are going to explode much higher. It’s a fun exercise to try to see roughly what is priced into the valuation in terms of growth, and picture what the world looks like with that amount of AI revenue in 5-10 years.

[Fernando Vidal @fernavid]

Why would the stock go up 10x though? Assuming it’s valued appropriately today, normal equity-like (rf + erp ~= 10%) returns would get you reasonable multiple compression over time as long as revenue ramps in line with expectations wouldn’t it? The stock would go up more or less than that 10% each year only based on how far away the true revenue growth curve is from this base case path of course. Magical Claude illustration of the required revenue growth path attached for critique.

[Warren Pies @WarrenPies]

GPU availability continues to plummet.

New multi-year lows in availability for: B200, H100, and A100s.

Good luck getting a B200 (availability <5%).

Oil getting all the headlines (rightly), but, for markets, the dominant secular story is surging compute demand.

[Fernando Vidal @fernavid]

@emollick Seedance 2 being frontier on videogen (at least publicly released frontier) and being closed weights is a data point in that direction.

[Fernando Vidal @fernavid]

This is more likely a sign of capacity limitations under massive growth than a failure in engineering. They always seem to have issues when the (human) workday is kicking off on the east coast.

[Gergely Orosz @GergelyOrosz]

On one end, the Anthropic team is a massive user of AI to write code (80%+ of all code deployed is written by Claude Code). They ship amazingly fast.

On the other hand, seeing these beyond terrible reliability numbers suggests there might be a downside to all this speed:

[Fernando Vidal @fernavid]

@jxnlco @dwr What makes it better?

[Fernando Vidal @fernavid]

After that quick small bump higher immediately after earnings, we are at almost no change after-hours on $NVDA, the quietest AH reaction we've seen over the last 3 years.

[Fernando Vidal @fernavid]

@NickTimiraos @WarrenPies The effect is also visible in sectors that did not overhire in 2021. In our report we measured residuals against trendlines that excluded 2020/21 and found negative residuals even in AI exposed sectors outside tech.

[Warren Pies @WarrenPies]

GPU availability continues to plunge.

B200 at new lows...A100, H100 approaching multi-year lows.

Demand ramping…B200 rarely available in our data

[Alex Corrino @AlexCorrino]

Impressive! Basically everything in your post is wrong!

The few H100s that are being sold for $6k on eBay are “parts-only” meaning they don’t function, they are only for repairing other H100s.

In reality, used H100s are still selling for $15-20k+. Not bad considering the original MSRP was $25-40k depending on the exact variant!

What’s more, NVDA still makes H100s and sells them for nearly the same price it did at launch.

Lastly, I’ve included a screenshot of the H100 rental price index showing that H100 rental prices are actually growing again. We don’t have to use your flimsy math to determine their value, we can just check the market!

[Warren Pies @WarrenPies]

Nice study here from Jacob

Put together w/ the help of Caliban.