Compound AI (@getcompoundai) header
Compound AI (@getcompoundai) profile picture

The AI Analyst for Finance that you can trust

Joined September 2025
Photo by Compound AI (@getcompoundai)Photo by Compound AI (@getcompoundai)Photo by Compound AI (@getcompoundai)Photo by Compound AI (@getcompoundai)Photo by Compound AI (@getcompoundai)Photo by Compound AI (@getcompoundai)Photo by Compound AI (@getcompoundai)Photo by Compound AI (@getcompoundai)Photo by Compound AI (@getcompoundai)Photo by Compound AI (@getcompoundai)
Introducing Compound V2: moving finance beyond chats to autonomous, long-running analysts: Multiplayer: The first AI analyst that collaborates like a deal team member. Chat in a shared agent channel or over email, and the agent will take feedback from everyone before acting. Agentic Office: No more subpar packages or brittle desktop software. We rebuilt Office suite, so Compound’s cloud agents now run with near-perfect parity, and 4X+ faster performance for large Excel tasks. Extreme Scale: Delegate to one agent at a time, but in the background, the agent is orchestrating hundreds of subagents. This enables extreme asks that run 100+ tasks in parallel. Try now at getcompound.ai Comment “V2” for a month free on our Pro tier, and see below what you can do with Compound, that you can’t with Claude/ChatGPT/others.
55
7
9
91
67,049
Compound AI retweeted
AI in Finance is inflecting, and there is a serious power law, as in coding. Not everyone has figured out how to leverage agents fully, but when it happens you see an explosion in usage. A lot of it has to do with tooling. Those who can leverage many parallel, asynchronous agents running long-running tasks are getting crazy gains in productivity and the gap is widening. They are generating orders of magnitude more output, but are doing it more efficiently. While Chat is like sucking AI through a straw, Compound V2 is like opening the floodgates of productivity.
3
1
2
4
1,085
Anthropic's IPO at $78/share (top of range, book oversubscribed) = $1.40tn fully diluted. We ran management's numbers and the main forecasts on X through one model, so you don’t have to. FY30E revenue / adj. op margin → IRR at $78: Damodaran: $385bn / 24% → 8.8% FundaAI: $438bn / 26% → 14.0% Compound Base: $446bn / 29% → 19.7% Artemis: $681bn / 28% → 31.4% Mgmt (Goldman rev): $600bn / 36% → 38.0% SemiAnalysis: $946bn / 35% → 57.6% Download our dynamic ANT IPO Model + deck for free and plug in your own case: getcompound.ai/blog/anthropi…
1
1
3
6
8,595
4/ Only Damodaran and our Downside value it below $78; our Base is $96–101. The key risk is $452bn of compute commitments due before 2031 that don't flex if growth stalls.
1
1
214
Introducing IC Deck Bench, the private equity industry's first benchmark for Powerpoints that meet the bar for your investment committee. Compound remains at the efficient frontier, performing better than any native model. Compared to Opus 5.5 Max, Compound is ~3X faster at ~1/3 of the cost. Read more: getcompound.ai/blog/introduc…
3
427
We mapped 68 GW of US AI data center capacity against permits, state pauses and local restrictions after the midterms. Only about a third has cleared every gate. Most of what's stuck is waiting on Texas regulators, not local bans. Full tracker: getcompound.ai/blog/data-cen…
1
2
3
793
The backstops have the same gap. Google's support at Cipher and TeraWulf starts at rent or lease commencement, and a delay termination doesn't trigger it. Nvidia's $105bn PORTS-Pike guarantee is built the same way. Until rent starts, lenders hold developer credit.
1
77
How long does the Texas freeze last? Base case (SemiAnalysis): it lifts after ERCOT's Dec 10 report, a ~3-month slip. Bear case: it holds until the legislature (Jan 12 to May 31) passes new laws, 5-9 months more. Most 2027 capacity is past the gate. The big risk is 2028 phases.
70
Introducing PE Bench, the first benchmark for AI agents on real private equity work: diligence, investment models, IC memos and decks across 5 data rooms. Compound scores above the efficient frontier set by Anthropic and OpenAI. Benchmark here: getcompound.ai/pe-bench A few more examples on how we differentiate below 👇
2
1
3
8
1,933
5. Why Compound scores higher: Our harness routes each part of a task to the frontier or custom model best suited to it, and we test it continuously against our own finance benchmarks across PE tasks. Compound Office also makes Excel, Word and PowerPoint directly (no brittle open source tools). There's also a lot of other improvements we make under the hood on cost & speed, to make this the preferred AI platform for Private Equity.
2
2
59