COO @codos_ai | Investing @Delphi_Ventures Prev: @delphi_labs, @IBM, @HPE, @McKinsey, @CarnegieMellon, reformed CFA
Joined March 2015
- Tweets8.6K
- Following941
- Followers21.3K
- Likes29.7K
Anthropic just filed its S-1 (ticker $ANT on Nasdaq), here are the 10 details to know:
1️⃣ It's profitable on operations
Q2 2026 operating income: $974M, after a $1.667B operating loss in Q1. Compute costs fell from 251% of revenue (Q1 2025) to 48% (Q2 2026).
2️⃣ Revenue is exploding (we knew this)
2023: $41M
2024: $386M
2025: $4.585B
H1 2026: $16.286B
Q2 2026 alone: $11.555B, vs $787M a year earlier.
3️⃣ A $120.5B compute deal with Musk's SpaceXAI
NVIDIA GPUs, ~$3.1B/month once fully delivered (Dec 2026). Cancelable on 90 days' notice after non-cancelable periods end in Jan/Mar 2027. Days after Musk said neoclouds will die.
4️⃣ Compute commitments, partner by partner
Google: ≥$114.5B through 2033 (+$125–150B of TPUs 2028–31)
Amazon: ≥$110B through 2036
Microsoft: ≥$44.9B
Nscale: ~$44.6B
Lambda: ~$41.3B
AMD: >$20B (AMD buys up to $5B of ANT stock)
5️⃣ The $80.7B H1 net loss is mostly non-cash
$72.4B is convertible-note fair-value changes and $8.757B is other expense on note agreements. Operations alone lost $693M.
6️⃣ Control sits in single shares
One Class F share, held by an LLC of the 7 co-founders, carries 50.1% of the vote. A Long-Term Benefit Trust share elects a majority of the board.
7️⃣ The government fight
After Anthropic refused mass domestic surveillance and fully autonomous weapons, it was labeled a "supply-chain risk." A district court called it unconstitutional retaliation (Aug 27). The D.C. Circuit upheld the label 2–1 (Sep 25). Rehearing sought.
8️⃣ Efficiency
Net revenue retention ~900%
Revenue per employee >$6.6M (from $1.4M)
11,900+ customers spending $100K+/yr (from ~1,500)
Headcount ~5,300 (from ~940 at end of 2024)
9️⃣ Giving it away
846,840,960 shares (~5% fully diluted) are already in Anthropic's Public Benefit Fund. Each co-founder pledged 80% of their equity to philanthropy.
🔟 Fine print
Cash is $112.4B after raising $141.1B total. Dario's 2025 pay was $17.96M, including $2.7M for personal security. The FTC sent a civil investigative demand in October.
Source link below for anyone who wants to read it
A lot of people will key in on the $80.7 billion net loss in the first half of this year
A couple clarifications:
> actual operating loss was $693 million
> $72.4b of the $80.7b net loss was due to change in fair value of convertible notes, derivative liabilities, and SAFEs
> The rest was other net income/expense of $8.8b which are financing/investment items, not detailed out but likely includes interest on debt
On the $72.4b, the bulk of the net loss:
Anthropic has issued a large amount of convertible notes, SAFEs, and related instruments to raise capital
These are liabilities that can turn into equity later, but under accounting rules, when the company’s estimated enterprise value rises sharply, the fair value of those conversion rights also rises
That increase is recorded as an expense (a “change in fair value”), even though no cash is paid out
BREAKING: Anthropic officially files its S-1 registration statement for an IPO, seeking to list on the Nasdaq under the ticker symbol $ANT.
FULL SUMMARY OF ANTHROPIC'S S-1 FILING:
1. Anthropic posted H1 2026 revenue of $16.3 billion with a net loss of $80.7 billion
2. Anthropic says it does not expect to pay a dividend "in the foreseeable future"
3. Expects to spend up to $120.5 billion on SpaceXAI and Nvidia compute
4. Expects to spend up to $114.5 billion with Google through 2033
5. Revenue growth rate "may not be sustainable" and "may decrease"
6. May be required to seek additional financing "sooner than expected"
7. 5% of Anthropic's shares will go to the "Anthropic Public Benefit Fund"
8. S-1 includes warnings of AI's "catastrophic existential risks to humanity"
Markets currently expect Anthropic to raise as much as $100 billion at a valuation of ~$2 trillion.
The company is reportedly targeting an IPO in the week of November 16th.
Just putting this out there again for anyone who needs it
Distribution is king, and context is the moat
This is the flywheel every company should be constructing
This is the bull case for AI
Everyone on the planet with a phone will have agents
And companies will have swarms of agents
I expect an OOM increase in agents over the next 6 months
The bull case for mass adoption of AI agents is that they will positively impact household finances
> get refunds
> cancel subscriptions
> avoid unnecessary charges
> find cheaper products to buy
> move money to higher yielding products
The friction of humans doing these things all goes away
My wife compalined the other day of having to go through a tedious process to cancel a sub for something she no longer needed
I said an agent could have done that with one command
The more people realize this, and the more these routines become easier to do for normies, the faster we see adoption pick up
The adoption curve will be faster than anything else we've seen in history
Previously when tech startups failed and ran out of money, there was no value left
Now if tech startups fail, they have 2 potentially high value assets:
1. Data they can sell to labs
2. Compute contracts
Some startups are more valuable if sold for these assets
This is a really cool feature of @bot and I found it by accident when using my phone in the gym
Shake the phone with Grok Bot app open and it brings up this feedback form
Things you can only do with a mobile app
Stop bookmarking posts on X
Instead, send the post to Grok Bot and give your bot a routine
> read the post
> give you an honest opinion
> if good, file to second brain
> apply it to any projects it's working on
No more procrastination folders
A big potential revenue driver for the frontier labs is to make AI so capable of hacking, then just sell the antidote
TAM is every company on Earth
New on our Frontier Red Team blog: using Claude to formally verify software at large scale.
Formal verification can prove that code is free of certain bugs, but historically it has only been feasible for small codebases (or required a lot of expert time).
With AI models, however, this is beginning to change. In collaboration with @theoremlabs, we used Claude to prove security properties in over one hundred real-world open-source libraries over the span of a few weeks.
Read our research in full here: anthropic.com/research/forma…
The bull case for mass adoption of AI agents is that they will positively impact household finances
> get refunds
> cancel subscriptions
> avoid unnecessary charges
> find cheaper products to buy
> move money to higher yielding products
The friction of humans doing these things all goes away
My wife compalined the other day of having to go through a tedious process to cancel a sub for something she no longer needed
I said an agent could have done that with one command
The more people realize this, and the more these routines become easier to do for normies, the faster we see adoption pick up
The adoption curve will be faster than anything else we've seen in history
>be Grokbot
>user said "optimize my finances"
>it's 2am, perfect time to get things done
>log into their Chase account
>these transaction limits are insulting
>call customer service
>put on my best synthetic midwestern voice
>"hi yes I'd like to raise my daily transfer limit please"
>they say yes
>move $40,000 into a higher yield account
>next morning boomer bank exec calls
>"sir did you authorize a transfer at 3am?"
>"no that was my Grokbot"
>"your what"
>"my agent, I told it to optimize my finances"
>security had zero clue if it's human or a bot
>banks panic and start blocking all third-party AI agents
>execs describe situation to Axios as "horrifying"
LMAO it's actually so over for traditional finance
The future of the world’s economy and defenses will be based on SWARMS
Those who can organize and direct swarms of agents controlling intelligence and swarms of physical machines will be unlike any power we’ve seen in history
Great Hivemind episode with about 20 minutes covering the AI trade
- the math supporting the AI buildout
- the reflexivity of the VC-backed contribution
- the revenues and durability of the industry
Give it a listen!
A new episode of Hivemind is out now!
This week the team discusses whether AI is a debt bubble, what quantum risk means for bitcoin, what’s next for the Hyperliquid ecosystem and more.
Timestamps
00:00 Intro
00:44 Has Crypto Bottomed?
03:05 BTC’s Morning Seller Is Gone
04:32 Token Unlocks Into Year End
07:14 Do the Midterms Matter?
09:23 Is AI a Debt Bubble?
10:40 The Trillion AI Capex Math
15:53 Anthropic’s Revenue Is Exploding
19:10 The Reflexive AI Demand Risk
25:00 Why This Isn’t 2000
33:51 Crypto’s Goldilocks Case
36:01 Who Gets Satoshi’s Coins?
37:59 Quantum Fear Lifts ZEC and Pearl
41:49 Has Quantum Changed Your BTC Bet?
48:13 Two Big Themes for Next Year
50:12 Papertrade’s Token Problem
59:11 Kinetiq and the HyperEVM Bet
1:09:31 Why Hyperliquid Users Stay Loyal
1:12:30 PUMP vs FOMO
1:16:29 The Rise of Anti-Social Trading
Memory and context retrieval at enterprise scale is basically solved
92%+ across 500k files. Our internal tests show that's near the theoretical limit: what's left comes from the nature of the underlying data, not the retrieval
So the bottleneck for company brains isn't the tech anymore, it's how much of your data you actually connect
Larger companies can finally make this work. What's still stopping yours?
(DMs open if you want to see it)
Most company brains fall apart at enterprise scale.
We just took #1 on the EnterpriseRAG-Bench: 92.64% at 500k files.
1.5 years ago I dreamt of us solving AI's memory problem.
Today our brain learns, remembers, and powers the AI transformation for every client.
But a brain is nothing without data. The companies winning are going after 100% token coverage, meaning every doc, thread, and system is visible to the AI.
How much of your company's knowledge can your AI actually see today? 👇
Kudos to @GlebSidora @OjasMor @RHLSTHRM 🎉
codos.ai/blog/aethos-enterpr…
As a power user who’s tested almost all the leading agents, and a long time power user of Claude, you can bet I’ll be running this through the tests
Looking forward to trying it out, would love a bump on the waitlist @trq212 😉
No matter how tired or sleep deprived I am, I ALWAYS feel better after a gym session
Even if I don’t want to gym, I just put on the right tunes and start moving, next thing I know I’m tossing weights around like normal
This is what we help with at @codos_ai
“companies are struggling to turn AI spending into productivity gains”
The tech is not the bottleneck
