More 2026 trends: Solar costs, oil oversupply, and the startup slump

This is a continuation rather than a self-contained episode. Kann opens by asking to get off the data center subject, immediately concedes that may be impossible, and settles for treating it as a second-order effect; the rest is a tour of numbered slides. Bullard’s stated purpose for the deck is the closest thing to a thesis: regardless of the rhetoric, here is how it actually played out in the data, which is neither as grim as one camp would have you believe nor as triumphant as the other. The opening pair of slides shows what he means. Clean energy equities returned about twice what the broad US indices did in calendar 2025, which he reads as evidence that activity in these markets continues whatever the politics sound like. In the same year, US energy startup investment fell to roughly a quarter of its 2022 level, and from about 7% of all US startup investment to two and a half percent. Public and private capital were pointing opposite ways at once.

The solar segment is the one Kann calls a soapbox, and it carries the clearest reasoning turn in the episode. Module prices keep falling and are cheaper than ever, and yet installed system prices in the US moved the other way: residential down only 3%, commercial up 9%, utility-scale fixed and tracking up 9 to 10%. The cause is everything that is not the module. Engineering, planning and permitting, labor and tariff-inflated equipment prices sit outside the module manufacturer’s business entirely, and Bullard’s sharper point is that they sit outside the developer’s control too, since no amount of development skill changes the landed price of imported equipment. The transformer slide makes the same argument from the supply side. Prices have not come down and if anything are still ticking up, and his explanation is manufacturer hesitancy rather than indifference: nobody wants to commit to inflexible manufacturing capacity while future demand is this uncertain, which is the logic he also sees in gas turbines. Kann pushes back that transformers should be the easier case, being less capital intensive and less of an oligopoly than turbines, and predicts that companies will come solve it. They agree that so far nobody has.

Two of the remaining segments describe places the American conversation is not looking. Oil is comfortably oversupplied, by roughly 5% on the IEA’s start-of-year estimate, which is enough to damp prices and to explain the lack of enthusiasm for re-engaging with ultra heavy sour crude that would need something like a hundred billion dollars of infrastructure investment before it could flow again. Demand is higher than ever and the market still is not screaming for new barrels. Meanwhile Australia was installing about 40,000 residential storage systems a month in the second half of 2025, mostly as add-ons to rooftop solar that households already had, after a government program opened in July; Kann’s arithmetic is that a year at that pace would put batteries on something like 5% of Australian households. And China accounted for about 42% of global passenger vehicle manufacturing in the first three quarters of 2025, a higher share than Japan ever reached in its heyday, while Chinese electric medium and heavy duty truck sales went from 3,000 in 2020 to 81,500 in the first half of 2025.

The forward-looking material is explicitly speculative and the episode marks it as such. Bullard predicts US energy startup investment above $3 billion this year; Kann takes the over hard, guessing at a tripling off the $2 billion base while allowing that he does not expect a return to $8 billion. His reasoning is that energy has been elevated by the conversations around AI and sovereignty and that interest is now coming from a wider and more diverse set of investors. Bullard’s caveat is aimed at his own series: what counts as an energy company is about to get blurry as firms discover or pivot toward energy applications, which he thinks is a good thing and which would also make the number harder to compare year over year. The episode closes on two quotes he picked for the same reason. A Virginia-focused energy economist said in December 2024 that his 2032 data center load forecast was probably off by a factor of two without knowing in which direction, and the head of Instagram said in February 2025 that the industry may need drastically more or drastically less capacity than it thought to build frontier models. Bullard reads both as genuine rather than performative modesty, and Kann lands the plane on the deck’s recurring title: nobody knows anything.