prefshares

Curious.

[prefshares @PrefShares]

Guys, we should not be bidding $AVGO and $AMD higher on this news.

This is 1 step forward (maybe) for alternative silicon and 3 steps back for the buildout.

[prefshares @PrefShares]

“based on what we know”

The fact they are making this decision shows that they know things we don’t know.

[Evergreen Capital @evergreencap3]

I'm not defending Jensen. His actions were disgraceful.

But at the same time, based on what we know, it makes no sense for the board to fire him. Jensen is completely irreplaceable and $NVDA will suffer. If his behavior was shameful but not criminal, he should repent and remain.

[prefshares @PrefShares]

You know it’s so bad.

[prefshares @PrefShares]

Jensen is going to wake up a loser.

[prefshares @PrefShares]

Guys, we can't have this.

[Ara Kharazian @arakharazian]

NEW Ramp AI Index: In October, AI spend rose 2%. OpenAI and Anthropic growth fell to low single-digits.

Engineering teams drove 72% of spend, followed by customer service and leadership.

Given this relatively modest growth in spend, and new competitive dynamics that may stick around, the markt should look to new fronts in the AI trade.

[prefshares @PrefShares]

It must be really hard being a CEO of a company and telling your employees that your earnings were solid and the business is doing well but you just so happened to report earnings on a day when your stock’s factor was out of favor so you’re stock is down 2% but you outperformed your peers and if your factor would’ve been in favor that day you probably would’ve been +7%.

[prefshares @PrefShares]

I'm arguing two things.

(1) It's not clear your analysis holds if you're willing to spend signifigant resources optimizing the hardware for large-scale repetitive workloads (which all the hypers + labs do). Put simply – I think the hypers + neos can get similar tokens/W out of ASICs.

(2) That assume's there isn't also a long-term strategic imperative "tax" the largest buyers would be willing to pay. I believe there is a strategic imperative to diversify silicon sources and they'd accept slightly lower tokens/W to grow ASIC market share.

[prefshares @PrefShares]

$NVDA $AVGO $ANT solana:56DeqXiADWY2Y3iD8pdbpTi9jhXXneofTY5dp4Xjpump $AMZN $MSFT $GOOG $META

"In a land/power/shell-constrained world, GPUs win. If power is the binding constraint, you want the most tokens per watt, Nvidia gives you the most tokens per watt, and the chip’s price matters less when the MW is the scarce input."

The logic is incomplete.

Most ASIC capacity isn’t sold. It serves the owner’s own workloads: TPU for Google (and Anthropic), Trainium for AWS (most of Bedrock now runs on it), MTIA for Meta. OpenAI has its Broadcom chip, and Anthropic is standing up its own chip.

For those workloads, the question isn’t tokens/watt. It’s cost per useful token across every constraint. And:

(i) Power isn’t the only thing that’s scarce. Capital is too. When you’re spending essentially all your FCF on capex, paying Nvidia’s ~75% gross margin means less compute per dollar.

(ii) The gap is narrower than the bull case assumes. SemiAnalysis has TPUv7 as the cheapest FP8 inference option vs B200/B300 (https://t.co/dZRWLSFQqj), and Trainium3 holding its own even in training, historically its weak spot (https://t.co/0yOFBpNclI).

Add the strategic reasons (supply security, bargaining power, years of sunk software work) and the allocation logic is simple. Each scarce MW goes to ASICs first. Nvidia will fill what’s left.

A power shortage is a chance for ASICs to take share.

Nvidia is no longer the base supply. It’s the swing supply. (A fine place to be in a shortage, just a more cyclical one.)

[prefshares @PrefShares]

People aren’t against datacenters.

Being “anti-DC” is a dog-whistle.

People are against AI.

They are scared by it. They believe it is going to take their job and create a worse world for their children.

They believe it may usher in the end of the world.

This isn’t about “water use” or “electricity prices.”

[prefshares @PrefShares]

"But there are still gaps, and it is the reason my answer is that this is “almost” AGI. Most critically, the AI did not notice these problems itself. It did not see that the course was a slog or that it had tied itself in knots with its own rulebooks. This was not for lack of checking. Fable put critics, fact-checkers or judges on nearly every project, but those checks found errors and added fixes. None of them asked whether the work was too long, too repetitive, or no fun. That all fell on me, and if I hadn’t watched and taken action, the results would have often been less impressive (though not exactly a failure, either). Knowing when your own work is not good enough is a large part of what makes someone an expert, and it is the part Fable does least reliably."

[Ethan Mollick @emollick]

I had early access to Fable 5.2 for a bit & I think it crosses the line into being a sort of Jagged AGI. Here, I have it launch a business, do academic research, hack hardware, build 3D models & more. I think it is a good illustration of limits & powers: https://t.co/QiR0IFPrZY

[prefshares @PrefShares]

@WaterworldCapi1 Tell me

[prefshares @PrefShares]

No one knows.

No one knows how algorithmic and silicon improvements will drive effective compute supply per GW of capacity.

No one knows how much demand there will be.

No one knows what 2030 lab ARR will be or how mature the agentic internet and enterprise will be in a few years. We don’t even know what types of applications we will have.

No one knows what AI CapEx will be.

No one knows how many GWs we will need in 2032.

No one knows what WFE will be.

Anthropic does not know. Google does not know. SemiAnalysis does not know. Jensen does not know.

Nobody knows.

[prefshares @PrefShares]

@ChairliftCap I have no earthly idea

[prefshares @PrefShares]

Would be a great opportunity.

[Joey Brookhart @SaasquatchC]

I'm looking for someone to help our team cover the consumer side of AI markets. If you know anyone who's passionate on writing, researching, and modeling what's going on in consumer AI and legacy internet business models, please have them apply!

https://t.co/zWMqdxkBp0

[prefshares @PrefShares]

So the issue is that this is what she puts outs there as her “official position.”

But the “official position” is often a euphemism for informal positions.

“Giving Texans a seat at the table.”

How many city councils do you think are going to affirmatively vote for large DC projects?

[Wall St Engine @wallstengine]

Gina Hinojosa may not be as bad for Texas data centers as it looks.

Her plan asks for: no subsidies, union labor, no NDAs, and proof household power and water bills won't rise.

Much of that is already underway:

Amazon, Google, Meta, Microsoft, OpenAI, Oracle and xAI pledged in March to bring their own power and pay for transmission

Microsoft pledged to pay full property taxes without seeking tax breaks

Nvidia is working with Vistra, NextEra and Constellation on data centers that cut power use when the grid is strained

[prefshares @PrefShares]

I'm starting to be a real believer in AI. Will be a megatrend like no one is really imagining right now. I'm sure lots of ramps and drops along the way though

[prefshares @PrefShares]

Oh so you’re a bottleneck bro?

How about some hyperscalers for ya.

[SouthernValue @SouthernValue95]

@PrefShares The bottleneck moves to existing capacity. $AMZN $MSFT $GOOG have the most installed capacity and the most pipeline already permitted and powered.

[prefshares @PrefShares]

Do you like to make money?

[SouthernValue @SouthernValue95]

@PrefShares The bottleneck moves to existing capacity. $AMZN $MSFT $GOOG have the most installed capacity and the most pipeline already permitted and powered.