AI is trapped. Here’s what founders should build next

Core thesis (business opportunity)

The speaker argues the current AI era is “trapped” because the world has:

• Intelligence (AI models)
• But lacks:
  • Power to run it at scale (energy capacity)
  • Coordinated action / hands (agents that can act in the real world)

Best opportunity for founders: build the missing power and action layers instead of competing directly at the model level with hyperscalers.

---

Historical framework used to structure the opportunity

The speaker uses an analogy: Revolution requires 3 pillars:

1. Power (e.g., steam engine)
2. Intelligence (e.g., power loom)
3. Coordinated action (e.g., factory)

If one pillar is missing, growth stalls—used as today’s lens for AI constraints.

---

Strategy: “Go one step beneath hyperscalers”

Recommendation: don’t compete with hyperscalers at the model layer.

Instead:

• Rebuild the “power” and “materials” hyperscalers desperately need
• Let hyperscalers spend large sums while startups provide components they can’t build fast enough

---

Power as the immediate bottleneck (energy capacity)

The speaker cites a target:

• DOE: add 100 GW to the grid by 2030
• Framed as enough to power 16 more New York Cities

Why it matters: demand for AI compute and supporting infrastructure outstrips available grid capacity.

Three founder “entry lanes” into power

1. Short-term (software):

   • Squeeze more capacity from grids running at ~half capacity
   • Claimed market prize: ~$170B
   • Example: Gridcare (grid efficiency/optimization)

2. Medium-term (hardware + software): “base power”

   • Example concept: home battery fleets acting like one distributed grid
   • Example metric: “raising at a $12B valuation”

3. Long-term (new generation):

   • Small nuclear reactors near data centers
   • Geothermal
   • Eventual fusion (speaker suggests ~30 years)

---

“Intelligence needs hands”: physical AI / agents / robotics

In software, AI’s “hands” are agents (draft, decide, transact).

The next wave is physical AI: robots/autonomous machines that act in the real world—“atoms, not bits.”

Two lanes for physical AI founders

• Horizontal platform/tool (like GitHub/Stripe but for robots)

  • Biggest unlock: training data

• Vertical ownership (end-to-end) in a narrow but valuable physical domain

  • Positioning warning: “Generalist tools are good for everyone but great for no one.”

---

Enabling layer: raw materials as a new “company shape”

Core operational dependency claims:

• You can’t build data centers without steel
• You can’t move power without better conductors

Tactic: take traditional heavy-industrial processes (steel, mining, carbon fiber) and “rebuild like a tech company.”

Treat these industries as:

• Capacity-constrained supply chains needing modernization and scaling

---

What makes this revolution different (execution-oriented)

This isn’t framed as just “better tech,” but as:

• Who is building it and why
• Founders targeting energy, robots, materials

The speaker claims audacious visions become solvable because once intelligence exists, the remaining problems become engineering problems with clearer execution paths.

They also assert the “curve goes exponential again” due to the convergence of:

• Power + intelligence + action

And claim:

• Two pillars still don’t exist yet (implying execution gaps/opportunity).

---

KPIs / targets explicitly mentioned

• 100 GW by 2030 (DOE grid expansion target)
• Grid utilization framing: grids running at ~half capacity
• Market prize claim: $170B (software route)
• Example valuation: $12B (base power / hardware-software example)
• Fusion timeline (high-level): ~30 years

---

Concrete examples referenced

• Gridcare: pursuing “software that squeezes more out of a grid”
• “Base power” / distributed batteries: described via an example company with a $12B valuation
• Technologies/paths mentioned:
  • Home battery fleets (distributed energy as “one giant grid”)
  • Small nuclear near data centers
  • Geothermal
  • Fusion

---

Calls to action / investing posture (business execution emphasis)

The speaker frames investment interest in the missing execution pillars:

“two of those three pillars don’t exist yet”

Invitation/stance:

• Founders should “come talk to us” if they’re building teams focused on power, agents, robots, and materials.

---

Presenters / sources

• Presenter: not explicitly named in the provided subtitles (speaker begins with “Let me back up…” and ends with “come talk to us.”)
• Source mentioned: DOE (U.S. Department of Energy) for 100 GW by 2030