Follow/Set Notifications! AFPR.COM- Lehman Brothers, Raymond James, JackShu HF (since 1987) https://xitter.cf/t.co/Ogi49pYcGD
Joined July 2007
- Tweets50.7K
- Following1.7K
- Followers63.4K
- Likes5K
Pinned Tweet
$SPY $SPX $ES - My 2026 eoy roadmap/prediction. January starts out choppy (Vol up + Mrkt Up) but mostly we spin our wheels into Feb. With lows tested we grind up to a new ATH into May - where you sell and go away into a 14%ish drawdown - and all da bears come out calling for the gap below to fill (it doesn't) and we start a 2nd 1/2 rally (FOMO kicks in around midterms) ending the year up 15%ish -7853
$ANT From the Anthropic S1 $SMH $MSFT $NBIS Revenue: $41M (2023) → $386M (2024) → $4,585M (2025) → $16.3B in H1 2026 alone (+1,300% year over year). That's roughly 3.5x all of 2025 in six months.
GAAP operating loss has narrowed to just $(693)M for H1 2026; on the adjusted basis they claim, operating income is +$1.6B (10% margin).
$131.7B in cash and short-term investments at June 30, 2026 ($112.4B cash + $19.3B investments). They've raised $141.1B in aggregate.
Governance is locked down tight: the founders' LLC holds Class F stock with 50.1% of the vote, public Class A is capped at 49.9% in aggregate, and the Long-Term Benefit Trust elects a board majority through Class T. It's a controlled company, and a public benefit corporation on top of that.
Ticker ANT applied for on Nasdaq. But this is an initial filing — price range, share count, and net proceeds are all blank so far, so no valuation yet.
Surprises
The compute commitments are on a scale I've never seen in an S-1: Google ≥$114.5B, Amazon ≥$110B, Microsoft ≥$44.9B, plus ~$245B in new commitments signed after June 30 and an estimated $125–150B more in TPU compute. Google and Amazon are simultaneously suppliers, creditors, and (via their stakes) shareholders.
A $1.5B copyright settlement (Bartz) was charged in 2025 and court-approved in July 2026 — and it's disclosed alongside June 2026 Commerce export restrictions on their Fable 5 / Mythos 5 models, later lifted.
The odd items: a 5% public-benefit share donation, a $2.0B Q3 donation-matching charge, a $15B revolver upsized in September 2026, and disclosed "revenue-forsaken" usage policies (they walk away from revenue on safety grounds).
Net revenue retention is stated at ~900% — essentially every customer grows enormously.
The good
80% of revenue is consumption-based (API usage), not subscriptions; compute cost has fallen to 55% of revenue; they count 11,900+ customers at $100K+ and claim all of the Fortune 10 as customers.
$6.6M revenue per employee — absurd operating leverage for a company this size.
New models (Claude Opus 5.5, Fable 5.1) are claimed to cut inference cost materially, which is the whole margin story.
The bad / watch out
The headline GAAP net loss for H1 2026 is $(80.7)B — almost entirely non-cash fair-value marks ($72.4B on converts/SAFEs), so it's accounting, not burn. But it will dominate every headline.
The real risk is structural: take-or-pay compute commitments vs usage-based revenue. If demand stalls, the commitments don't.
Customer concentration, heavy US data-center exposure, sales and marketing spend spiking, a PBC structure with untested litigation exposure, and no Q3 financials in this filing — the June 30 cutoff is already stale given the post-quarter commitments.
Key numbers
Revenue: 2023 $41M · 2024 $386M · 2025 $4,585M · H1 2026 $16,286M
GAAP operating loss H1 2026: $(693)M · Adjusted operating income: +$1,618M (10%)
GAAP net loss H1 2026: $(80.7)B (non-cash marks)
Cash + short-term investments: $131.7B · Capital raised: $141.1B
Compute commitments: Google ≥$114.5B · Amazon ≥$110B · Microsoft ≥$44.9B · ~$245B post-June additions
Control: Founder LLC 50.1% (Class F) · Public Class A capped 49.9% · LTBT board majority (Class T) $GOOGL
$SPY Slack/Discord functionality coming soon....So you can easily search for the stock you want to know about...see the charts history, and the latest chart...EVERY channel (and chart) you can opt-in for alerts when it's updated - Sign up for the beta! afpr.com
Go to afpr.com and click on Tools-beta ...All subscribers (paid and free) to my substack will be granted immediate access....play with the 19 (more coming everyday) screeners there, give feedback...recommend one's YOU would want to have...
It's still pre-beta but this is gonna be killa!!!! ...all based on 1000's of charts and patterns i've identified over years...help me build this...feedback as to what to screen for appreciated...the RSI screen already blows away everything!!
Anni's great make sure you're following her $NBIS
Here was my case yesterday on FOX BUSINESS for miners and Neoclouds among the worst day for this group. Also my take on Micron $MU product and Tech Roadmap. foxbusiness.com/video/640724…
Not clear why people are drinking on Muse (ya all these guys will have the same thing). He just built my website (phase 1...phase 2 he will be replacing my slack and substack) afpr.com ...he also found me used snow tires mounted on prius rims for my son's car- and just now bought me a 2017 Outback for $1700 (I will flip it for 6K) $FB
$SPY $QQQ Working on a new indicator based on the dynamic MA cross....basically adds a box that doesnt add a million MA's onto the chart (the timeframes dynamic F/S ma's are shown by default but can be turned off in settings - The idea is you get a quick visual in an unobtrusive box that let's you know (it changes color and gives a confidence score based on how many MA's it is above) what MA's we are above or below...alerts you when you are approaching one (from above or below) and when you're touching one - includes VWAP and anchored VWAP!!
If you're investing in the space @IvanaSpear is a MUST follow (sub to her newsletter as well) $MSFT $NBIS
Investors are paying record prices to insure AI debt.
The market is worried about the buildout.
But it isn't pricing every builder the same.
Insuring $1M of CoreWeave's debt costs over $100,000 a year.
For Oracle, which S&P cut to one notch above junk in July, it's $31,400.
For Meta, it's $11,800.
The biggest builders carry some of the strongest credit ratings in the market.
S&P Global rates Microsoft AAA, Amazon AA and Meta AA-.
In February, S&P estimated Amazon, Meta and Alphabet Inc. could each borrow nearly $200B and keep their ratings.
So the companies with strong credit will keep building.
Weaker borrowers will find it harder, and pricier, to keep up.
That means a less exuberant buildout.
And a less exuberant buildout means higher returns for longer.
GPU rental rates are already rising: B200 prices rose 28% in the first eight months of 2026.
I expect them to keep climbing.
Follow @IvanaSpear, like this post, and comment "CREDIT"; and I'll send you this week's Tech Edge on who can fund the buildout.
$CRWV $ORCL $META $MSFT $AMZN $GOOG
$SPY find me someone who does it better GO!; afpr.com/examples
$CL Will the cat get out of the bag? ... $SMH What are the next levels to watch in the semi space? ...Join us in TGIF Chat today 100% free ...just subscribe (also free);
annualizethis.substack.com/c…
Also MiniMax... (All time lows)
MiniMax M3.1 Flash weights are now public. 428B parameters total, 23B activated per token.
Agentic performance: A generational leap in agentic tasks, coding, and cybersecurity — 76.2 on BankerToolBench, 73.7% on DeepSWE v1.1, 61.7 on OSWorld V2, 29.6 on Agents' Last Exam, and 82.2 on CyberGym.
Efficient by design: MiniMax Sparse Attention and DSpark speculative decoding improve inference efficiency.
Hugging Face:
huggingface.co/MiniMaxAI/Min…
ModelScope:
modelscope.cn/models/MiniMax…
API: platform.minimax.io
M Plan: minimax.io/m-planMiniMax
MCode for desktop: agent.minimax.io/download
