The AI Power Rankings: grading the people building the future

The AI boom stopped being a single story. By mid-2026 it's a field of named bets, each tied to a person whose decisions move trillions. Some are running away with the race; some are coasting on a narrative the numbers no longer support. I build agentic AI for a living and track these companies as both a paying customer and an analyst — what they charge me, and what the market charges to own them. Here's the report card.

Sam Altman — OpenAI

Performance. The most famous operator in the field, and on raw capital and reach he's delivered: a record $122B raise at an $852B valuation, revenue run-rate from ~$2B (2023) to ~$25B annualized by early 2026, ChatGPT still the default for hundreds of millions. He completed the nonprofit→PBC conversion (the Foundation keeps board-appointment rights and a ~$130B stake) and renegotiated Microsoft (capped revenue share, dropped exclusivity, multi-cloud).

Problem. He's barreling toward a giant IPO with potholes that have his name on them: Republican AGs and the House Oversight Committee are scrutinizing conflicts of interest. Worse, OpenAI lost the technical lead — Gemini 3 beat it badly enough to trigger an internal "code red," and it shut down Sora. Still deeply lossmaking.

Forecast. A generational fundraiser and distributor with a durable brand — going public having surrendered the performance crown, with regulators circling and a cost structure needing near-unlimited capital. Verdict: a strong franchise that's no longer the front-runner.

Elon Musk — SpaceX, xAI, Tesla

Performance. The most audacious financial engineering in corporate history: folded xAI and X into SpaceX (a $1.25T entity, largest merger ever) and took it public June 12 at ~$1.77T, raising ~$75B — the largest IPO ever, making him the first trillionaire. Underneath: a real moat — reusable launch ~94% cheaper per kg than China, Starlink 7,000+ satellites.

Problem. The valuation assumes that lead is permanent plus a speculative orbital-compute future on schedule; analysts say fair value is $600–900B. ~80% of his fortune rides on that one mark (~95% unrealized). And Tesla — ~350× earnings while BYD outsold it (2.26M vs 1.64M), out-earned it (~$118B vs ~$92B), and trades at a fraction — is worth <1/10th as much, shielded at home by 100% EV tariffs.

Forecast. SpaceX is the real asset; xAI's Grok is cheap but trails on capability; Tesla is the overhang. Verdict: dazzling capital engineering and a real space moat, married to a stretched valuation, extreme concentration, and a car company losing the cost war it pretends it isn't in.

Dario Amodei — Anthropic

Performance. The quiet winner: a $65B Series H (May 28) at $965B, past OpenAI to most valuable AI lab, on ~$47B annualized revenue. Claude (Opus 4.8, Fable 5) is the preferred stack for serious coding and enterprise — the highest-value, stickiest segment. Confidential October IPO filing; guiding to break-even by 2028.

Problem. Still lossmaking and capital-hungry, facing the same Chinese commoditization pressure as everyone; consumer presence is a fraction of ChatGPT's.

Forecast. The cleanest ascent in the field — passed OpenAI, owns the most defensible niche, carries a far less messy governance story into its IPO. Verdict: disciplined, focused, ascendant — the lab that quietly took the lead.

Sundar Pichai — Google / Alphabet

Performance. The comeback of the cycle, and on the numbers the best on this list. Alphabet is up ~75% on the year, crossed $4T (first time), passed Apple, and is near overtaking Nvidia. Q1 2026: $109.9B revenue (+22%), net income $62.6B (+81%), Cloud +48%. Gemini 3 outperformed every rival, crossed 750M MAU, and cut AI query cost ~78%. Apple pays Google ~$1B/yr to run a custom Gemini in Siri.

Problem. $175–185B of 2026 capex (a near-doubling) pressures margins, and only pays off if infrastructure keeps converting to demand; search antitrust is a live risk.

Forecast. The best-positioned full-stack player, full stop — owns the chips (TPUs), models (Gemini), distribution (Search, Android, Chrome, YouTube, now Siri), cloud, and data. As workloads shift to inference, cost-per-query is the game and owning silicon is the edge. Verdict: the standout — from afterthought to arguably the safest large-cap AI bet.

Jensen Huang — Nvidia

Performance. The undisputed financial winner: first company to $5T, trading in a $5–5.5T band no one else has touched. A recent quarter: ~$81B revenue (+85%), net income $58B (+211%). He sells the picks and shovels to everyone here — including those building chips to compete with him.

Problem. The customers are the competition (Google TPUs, Broadcom's $8B+ AI quarter, Marvell), especially on inference; the valuation requires the entire circular capex buildout to keep compounding. Any capex pause hits Nvidia first.

Forecast. Vera Rubin ramps H2 2026 and CUDA keeps Nvidia central; Huang is underwriting a path to $10T. Verdict: the cleanest winner of the cycle — but increasingly a leveraged bet on the AI capex cycle never slowing.

Liang Wenfeng — DeepSeek

Performance. The disruptor, and relative to his resources the most consequential person here. V4-Pro at $0.435/M input — ~22× cheaper than Opus, ~44× cheaper than Fable 5 — at ~90–95% of capability on public benchmarks, running on Huawei Ascend hardware, routing around US export controls entirely.

Forecast. Not a valuation story (privately backed, not chasing a trillion-dollar mark) — a margin story, and the asymmetric threat to every Western name above. By making "good enough" intelligence nearly free, he compresses the whole industry's economics and is living proof that export controls are accelerating, not preventing, Chinese self-sufficiency. Verdict: the highest disruption-per-dollar in the field — the man who made intelligence cheap.

The fast follower: Zuckerberg

Meta spends at hyperscaler capex pace and pursues an open-weight Llama strategy that, like DeepSeek, pressures closed-lab pricing from below, with enviable distribution. But it monetizes attention, not intelligence directly, so its AI spend is a defensive moat around advertising rather than a standalone engine. Watchable, not yet decisive.

The ranking

On the combination of what they've delivered and where they're positioned: 1) Pichai (execution + full-stack), 2) Huang (financial king, single point of failure = the capex cycle), 3) Amodei (disciplined new valuation leader), 4) Liang (the disruptor reshaping everyone's margins), 5) Altman (powerful but slipping, under a governance cloud), 6) Musk (not because SpaceX isn't real, but because the blended bet has the widest valuation-vs-fundamentals gap, the most concentration risk, and the weakest core asset in Tesla).

The through-line: the winners own infrastructure and distribution (Pichai, Huang) or made intelligence so cheap they reset the board (Liang). The most exposed are those whose valuations have outrun their economics and now depend on a narrative holding or a capex cycle never breaking. The IPO window — Anthropic in October, SpaceX already trading — is where those two groups get sorted in public, on real numbers, for the first time. I'd watch that repricing more closely than any model launch this year.

Bo Shang builds agentic AI and security tools, including the Erosolar research agent and the Trenchwork defensive-cyber platform. Reach him at bo@shang.software. Figures are mid-2026, from public reporting and company disclosures; verify against primary sources. Not investment advice.