Daan Tav

Building | Equity research for growth & tech stocks | $TSLA $AMD $GOOGL $PLTR $NBIS | Not financial advice

[Daan Tav @AlphaForgeDaan]

A thread on why $AMD might be the decade's most compelling AI investment 🧵

1/ The Market Opportunity: AI Is Rebuilding Computing

We’re witnessing the biggest infrastructure buildout in tech history.

Data centers are expected to expand from ~$200B → $1T by 2030.

The question isn't if we reach $1T in AI infrastructure spending.

The question is: can AMD capture 10-15% of it while expanding margins?

History suggests betting against Lisa Su's publicly stated targets has been... expensive.

[Daan Tav @AlphaForgeDaan]

@EndicottInvests Meanwhile Nebius just raised its H100 price from $3.85 to $4.50 an hour, about 17%. And that’s for a chip from 2022.
Long term outlook is solid imo.

[Daan Tav @AlphaForgeDaan]

$TSLA setting a capex record this quarter isn’t really a surprise, it���s exactly what management guided for.

The current record is Q2’s $5.79B, and Tesla guided 2026 capex above $25B. With ~$8.3B spent in the first half, Q3 and Q4 need to average over $8B each.

For scale: Tesla spent about $8.5B on capex in all of 2025.

[Kalshi @Kalshi]

JUST IN: Tesla's quarterly capex is set to hit a record on its physical-AI push, per Bloomberg

[Daan Tav @AlphaForgeDaan]

@GrindeOptions $5B might even be light. Q2 was already $5.79B, and to hit the $25B+ they guided for 2026, the second half needs to average over $8B a quarter.

For context: in all of 2025 $TSLA spent about $8.5B on capex, so this year alone is roughly 3x last year.

[Daan Tav @AlphaForgeDaan]

Helios was always the biggest risk in the $AMD story.

But the more important number in this KeyBanc note: AMD’s CoWoS supply for 2027 was revised up more than 30%.

CoWoS is TSMC’s packaging step that joins the GPU with its memory, and it’s been the main bottleneck for AI chips. That extra capacity means AMD’s supply chain is preparing for much bigger GPU volume next year.

This aligns with AMD’s own guide for data center revenue to more than double in 2027.

[P Equity Research 📰 @pequityresearch]

Keybanc: we hear that AMD's ODM partners are really struggling with ramping production volume of Helios racks as they are undergoing similar growing pains to what NVDA did when it initially ramped GB200. The two things that we have heard as causing the most issues are with the OAM (Open Accelerator Module) GPU board and cartridge connector issues (this is very similar to the issues that NVDA faced on GB). While we previously had expected 4K Helios racks this year, we now expect less than 1K racks. For next year, supply chain partners indicate expectations to ship 20K racks.

$AMD

[Daan Tav @AlphaForgeDaan]

@treasureh8nter Worth keeping an eye on margins through this too.

$AMD already guided MI450 below its company average gross margin during the ramp, and a bumpy ramp likely doesn't help that.

[Daan Tav @AlphaForgeDaan]

@pequityresearch Worth remembering the orders behind these racks are already signed, OpenAI and Meta for multiple generations. A slow ramp delays that revenue, it doesn’t lose it.

[Daan Tav @AlphaForgeDaan]

@zerohedge Plot twist: SoftBank/Jane Street as tenant trades wider than CoreWeave as tenant. 683 vs 502bp.

[Daan Tav @AlphaForgeDaan]

@garyblack00 Fair on the EPS math, but -3.1% vs -2.3% for the NDX isn’t much of a signal.

Tesla’s beta has historically been well above 1, so on a down day for the market that move is about what you’d expect anyway.
If anything, the event barely shows up in the price.

[Daan Tav @AlphaForgeDaan]

@gurgavin $TSLA product events have looked like this for years. Buildup before the reveal, selloff after, regardless of how impressive the product is.
Market is pricing in flying cars now as well.

[Daan Tav @AlphaForgeDaan]

@WheelieInvestor That’s been the pattern with $TSLA product events for years. The stock runs up into the reveal, then sells off once it’s out, even when the product delivers.

[Daan Tav @AlphaForgeDaan]

In January I wrote this thread on $AMD when it was around $209. It’s now around $542, already past the $500 I called my 2030 base case.

So, is it still a buy at this price?

Depends on which earnings you look at.

On what AMD earned over the last 12 months, it trades at ~139x (GAAP). Expensive by any measure.

On management’s own target of $20+ EPS by 2030, it’s ~27x (non-GAAP). Not cheap, but reasonable for a company growing this fast.

So at $542 you’re paying a premium today for earnings AMD still has to deliver. Most of that growth depends on Helios, its first full AI rack. AMD has never shipped one at volume before, and that’s where the biggest risk sits.

I’m still buying, just small. Below $485 it drops to ~24x, and that’s where I’d start buying in larger amounts.

[Daan Tav @AlphaForgeDaan]

A thread on why $AMD might be the decade's most compelling AI investment 🧵

1/ The Market Opportunity: AI Is Rebuilding Computing

We’re witnessing the biggest infrastructure buildout in tech history.

Data centers are expected to expand from ~$200B → $1T by 2030.

The question isn't if we reach $1T in AI infrastructure spending.

The question is: can AMD capture 10-15% of it while expanding margins?

History suggests betting against Lisa Su's publicly stated targets has been... expensive.

[Daan Tav @AlphaForgeDaan]

@TheValueTrade Approval isn’t really the main point here.
It’s about showing that something that looks impossible on paper can actually be built. Today it’s a flying car, next it’s FSD at global scale and humanoid robots.

[Daan Tav @AlphaForgeDaan]

Agree on the valuation caution, and the Roadster math makes sense. It’s a rounding error on Tesla's entire earnings.
Where I’d disagree is autonomy commoditization. FSD at Tesla’s cost and scale isn’t something anyone else is close to yet.
That could change eventually, but I don’t see it happening within the next few years.

[Daan Tav @AlphaForgeDaan]

@TheLongInvest Very impressive, but what I really want to know is what’s keeping it on that wall. If that’s the $SPCX hardware doing the work, the line between the two companies is getting thinner.
Makes the merger talk feel a little less far-fetched.

[Daan Tav @AlphaForgeDaan]

One product event doesn’t say much about where $TSLA goes long term. If anything, it’s another case of Tesla delivering what it announced, however crazy the announcement sounded at the time.

The Roadster was never a big part of the story financially anyway, autonomy and Robotaxi are. And the Nike comparison doesn’t really hold up for me. Nike’s problem was its core business losing ground.

[Daan Tav @AlphaForgeDaan]

No doubt it'll get a lot of attention, and I'll be glued to the screen tonight like everyone else.

Cybertruck got that kind of attention too though, and I'm not sure it ever clearly showed up in demand for the rest of the lineup. Would be curious to see if the Roadster does better on that front.

[Daan Tav @AlphaForgeDaan]

@garyblack00 @TSLAFanMtl Day-one reaction probably disappoints, like Gary says.

But the Roadster was never going to matter for the $TSLA share price on its own anyway. It's one more "Tesla delivers, timelines vary" moment, and those add up over years in a way a single day's move doesn't.

[Daan Tav @AlphaForgeDaan]

@TheLongInvest Same company at $534 as at $650, just a little more likeable now.
Still waiting for a level I'd actually add $AMD again at though.

[Daan Tav @AlphaForgeDaan]

So tonight, enjoy the show. Then write down the production date.

Each of those products eventually shipped. If the Roadster follows, it’s one more proof point for the bigger bets still ahead, robotaxi at scale and Optimus, where the stock valuation really sits.

Full breakdown in our Tesla Read: https://t.co/ZqZyOSswga

[Daan Tav @AlphaForgeDaan]

$TSLA Nearly nine years after it rolled out of the back of a Semi trailer, the Roadster is finally getting its moment.
Most people will be watching the 0-60 time. We’re watching one number: the production date.🧵

[Tesla @Tesla]

Roadster event streaming live this Thursday at 7:55pm CST

[Daan Tav @AlphaForgeDaan]

The Roadster itself won’t move the stock. Tesla just retired the Model S and X lines to make room for Optimus, so a low-volume sports car is not where the earnings inflection will come from.

But it’s a reminder of what Tesla’s record actually says: the promised dates slip, but the products arrive.