1.

Order books at GE, Siemens, and Mitsubishi stretch to 2029. Turbine prices have nearly tripled since 2019. Stone Power claims it can deliver gigawatts before 2030 by building an AI engineer that optimizes at the system level — not the component level — and then manufacturing the turbines itself. The pitch: Stone is both Nvidia and TSMC.

2.

GPU production is outrunning the grid

• The US needs 1.5 terawatts of new power by 2035 — a figure stated without source, but the author's own textile factory in North Carolina paid electricity bills approaching 30% of COGS
• Order books at GE, Siemens, and Mitsubishi stretch to 2029; turbine prices have nearly tripled since 2019
• Every AI data center needs power and every gas plant needs a turbine — yet one bottlenecked part holds up the entire supply chain

3.

Stone's AI engineer optimizes what no human can

• Turbomachinery physics has been studied for 120 years — enough ground truth to train against, but the trade space across vibration, mechanics, fluids, combustion, and thermodynamics is brutal enough that no human engineer can fully optimize it
• Modern simulation closes most of the gap to physical test; what's left is exactly the kind of global optimization problem AI is built to solve
• Stone's agentic toolset can 100x any hardware engineer — but the toolset only matters because the hardware does

4.

The incumbents are optimizing the wrong thing

• Stone's first product is a large natural gas turbine optimized for total power volume to market, designed from first principles around the supply chain failures crippling the OEMs
• The mistake everyone else is making — incumbents and startups alike — is getting nerdsniped into simple cycle efficiency and locally optimizing components without manufacturing in mind
• Stone optimizes at the system level, simplifies radically, and claims it delivers years faster

5.

Stone is both Nvidia and TSMC — design and manufacturing under one roof

• Stone needs to make thousands of turbines and other hardware to gather manufacturing, testing, materials, service, and deployment data at sufficient scale to inform hundreds of custom designs
• The data is only half the moat; the ability to make designs into atoms rapidly closes the loop
• Stone will build its own factories and deliver its own natural gas, steam, and advanced turbines — a vertically integrated model the author claims no competitor has

6.

Natural gas is the first leaf on a multi-decade tech tree

• The same physics kernel, the same factory, and the same engineer can deliver nuclear turbines, geothermal turbines, heat recovery systems, and propulsion engines
• Dozens of companies are working to unlock energy abundance through splitting the atom — few have figured out how to convert heat energy to electrical power at scale
• Stone's mission: deliver natural gas turbines now to build the AI design stack and the world's best turbine factory, then use it to enable nuclear companies to deliver the new atomic age

7.

The entire pitch rests on a turbine that has never been built

• Stone Power has raised a preseed from Boost VC and Forward Deployed VC — the earliest stage of venture funding — and has disclosed zero technical specifications, zero test results, and zero customer contracts
• The claim to deliver gigawatts before 2030 is stated as fact but supported by nothing; the $200B+ market figure is asserted without source; the 100x engineer productivity gain is a slogan, not a measurement
• The "both Nvidia and TSMC" analogy is aspirational, not descriptive — Nvidia spent decades building its design moat, TSMC spent decades building its manufacturing moat, and Stone Power has spent less than one year in stealth with a preseed check

8.

The real bet isn't on a turbine — it's on a design loop that compounds

• Both the original pitch and the counter-argument assume the turbine is the product; it isn't — the product is the data-driven design-manufacture-test loop that improves with every turbine Stone builds
• Nvidia's moat isn't any single GPU; it's the compounding cycle of design, fabrication, and feedback across thousands of chips — Stone is attempting the same flywheel in a market where incumbents have no incentive to move at startup speed
• If the loop works, Stone doesn't need to beat GE on any single turbine spec — it needs to iterate faster than GE can react, and in a supply-constrained market where order books stretch to 2029, the incumbents have already surrendered the speed advantage