Feb 6, 2025 · 48m · catalyst
More 2025 trends: DeepSeek, plug-in hybrids, and curtailment
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
Host Shayle Kann and energy analyst Nat Bullard evaluate key data trends shaping the global energy landscape, analyzing renewable curtailment, transmission stagnation, and power load growth. They explore how breakthrough AI models like DeepSeek and the global surge in Chinese extended-range electric vehicles are redefining infrastructure and technology markets.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Shayle holds 31.6% of the talking time here. How this is scored →
speaking balance: gold is Shayle, purple is the guest (3 minute bins)
Nat sharply corrects the framing around historical load growth forecasts, emphasizing that past forecasts anticipated modest growth but realized actual absolute zero growth.
Hardest push from Shayle ▶ 19:41 Historical load growth contextualizationShayle resists the prevailing panic over AI load growth, offering counter-analysis that 1.5% annual growth is still below early 2000s 2% benchmarks.
Biggest teaching moment ▶ 45:51 Deconstructing AI training cost breakdownNat educates listeners and Shayle on Epoch AI empirical data demonstrating that electricity makes up less than 7% of AI training costs, far behind talent and equity.
Shayle holds their own ▶ 23:29 Jevons paradox synthesis for AI computeShayle synthesizes market reactions to DeepSeek by laying out the dual economic scenarios of pure power reduction versus Jevons paradox-driven demand surge.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Shayle as informed peer | Guest teaching | Guest disagreement | Shayle pushing back | Why |
|---|---|---|---|---|---|---|
| Host Welcome and Overview of Part Two | 6 | 5 | 1 | 2 | Shayle opens with a nuanced take comparing US solar and wind trajectories against global records, noting policy headwinds under new administrations. Nat elaborates with Rhodium Group and ERCOT deployment data showing storage eclipsing wind investment. The dynamic is highly collaborative and analytical. | |
| Renewable Curtailment and Negative Pricing in Grids | 7 | 5 | 1 | 2 | Shayle cites California's negative spring net load and massive curtailment despite perceived grid supply shortages. Nat explains the four-hour storage regulatory cap and contrasts it with Europe's 9,000+ negative pricing hours. Both demonstrate deep fluency in electricity market distortions. | |
| Grid-Scale Battery Performance in Extreme Temperatures | 5 | 6 | 1 | 1 | Shayle frames the reliability question comparing thermal capacity value against battery availability during weather extremes. Nat presents Modo data showing ERCOT grid-scale batteries maintaining mid-to-high 90% availability even in sub-freezing or 105F+ temperatures. | |
| Mid-Roll Sponsor Messages: Bloom Energy, Engie, and EnergyHub | 7 | 5 | 2 | 3 | Following mid-roll ads, Nat recounts hand-collecting NERC reliability forecasts to show the unwinding of historical load stagnation. Shayle pushes back on the panic by placing current 1.5% load growth projections into context of early 2000s 2% growth rates. Nat reframes that earlier forecasts predicted growth where actual realization was flat. | |
| DeepSeek's AI Breakthrough and Computing Scale | 8 | 6 | 2 | 3 | Nat details DeepSeek V3's dramatic efficiency (2.8M training hours vs Llama's 31M) and open-source economics. Shayle counters with a clear articulation of Jevons paradox versus reduced capacity requirements, and explores the electrical siting implications of 20MW distributed clusters versus gigawatt campuses. | |
| The Severe Stagnation in US Transmission Construction | 6 | 4 | 2 | 2 | Shayle highlights the severe stagnation in US high-voltage transmission, prompting Nat to contrast 2013's 3,200 miles with 2024's 125 miles built. Both agree that the failure is rooted in permitting and jurisdictional gridlock rather than capital constraints. | |
| The Global Rise of Chinese Extended-Range Electric Vehicles | 5 | 7 | 1 | 1 | Nat educates Shayle on modern Chinese extended-range electric vehicles (EREVs) with 200km battery range and small generators, challenging Western perceptions of legacy plug-in hybrids. He shares firsthand observations from Southeast Asia and Australia of Chinese EV utility vehicles capturing market share. | |
| China's Dominance in Global Automotive Manufacturing and Exports | 6 | 7 | 1 | 1 | Nat presents macro statistics showing China exporting 6.4 million vehicles annually and controlling nearly 40% of global vehicle production, while foreign joint ventures like GM collapse domestically. He illustrates this with BYD surging from 0.2% to 14.4% market share in Singapore within three years. | |
| Data Center Capital Expenditure Surpasses Hospital Construction | 6 | 5 | 1 | 1 | Shayle flags that data center building capex ($31B) has overtaken hospital construction ($27B). Nat explains the structural shift from on-premise server rooms to hyperscale facilities and raises whether lightweight open-source models could reverse the centralization trend. | |
| Regional Concentration, Duty to Serve, and Stranded Asset Risks | 7 | 7 | 1 | 2 | Shayle highlights Northern Virginia data center power consuming 26% of state load and notes absurd interconnection requests in small electric cooperatives. Nat explains the regulatory crisis around the utility 'duty to serve' and warns of stranded asset risks for ratepayers if speculative gigawatt requests vanish. | |
| The Bit-Watt Spread and AI Training Energy Cost Dynamics | 7 | 7 | 1 | 1 | Shayle introduces the 'bit-watt spread' concept to examine willingness-to-pay dynamics. Nat walks through Epoch AI data showing energy comprises only 1.7% to 6.3% of model training costs compared to massive human capital and equity costs, illustrating why tech buyers are price-insensitive compared to aluminum or steel manufacturers. |