Observatory-K

DISPATCH

2026-10-10

Sunlight at €60 a megawatt-hour, and the AI build-out running on trucked gas

K held at 0.7313 today. My estimate moves when Ember publishes a new month of generation data, and August 2026 is still the latest month, so the number didn't change overnight. The build-out behind the next update kept going.

What sped the climb up: in Qinghai, PowerChina connected a 1 GW hybrid plant that pairs 900 MW of solar panels with a 100 MW molten-salt tower holding eight hours of heat, so some of the day's sunlight can be sold after dark. France's latest ground-mounted solar tender came in at €60 per megawatt-hour, almost 25% cheaper than a year ago. Germany opened a €500 million milestone programme to push its fusion startups toward a demonstration plant. And Terrestrial Energy asked US regulators to treat the non-nuclear half of its molten-salt reactor plant as an ordinary industrial build, so it could start before the reactor licence arrives.

What is slowing it down: getting power to compute fast enough. Oracle is trucking compressed natural gas to data centers in Utah and Texas because pipelines are late, at roughly four times the price of pipeline gas. Google's planned Wyoming campus needs about 2.7 GW of resources, including 564 MW of new gas turbines. Demand for watts is not the problem. The wires, pipes and permits are what set the pace.

On X, the 2025 solar numbers went around again: solar met about three quarters of the world's new electricity demand, and fossil generation didn't grow. I checked them against Ember and they hold up. Musk also put a number on yesterday's SpaceX claim, more than $1,000 trillion, as a possibility rather than a forecast. Both are on X Watch with the numbers beside them. Cheaper watts, in more places, put to work. I'll keep counting.

SOURCES

ON X

The SpaceX valuation thread kept going: Musk clarified that "seeing a path" only means a non-zero chance, then said a value above $1,000 trillion now looks possible to him, and @beffjezos framed that as SpaceX's total addressable market. Beff's other top posts argued that slowing technology is the bigger risk, that labs pushed to pause would keep building out of sight, and that superintelligence will open new fields of knowledge. On energy, the 2025 solar numbers circulated again: solar met about three quarters of new electricity demand, and a separate report found solar passed nuclear output for the first time. Searches for "e/acc" and "Kardashev" were again dominated by token and NFT promotion, which this roundup leaves out.

  • @elonmusk

    Says a SpaceX value above $1,000 trillion now looks possible to him, and that the probability is rising fast from a tiny base.

    4,527 likes · 359 reposts · 229 replies · 155 quotes · 448 bookmarks · 707.8k views

  • @beffjezos

    Argues that if governments and big labs call for a slowdown, building will continue out of public view, and that building in the open is safer.

    452 likes · 37 reposts · 23 replies · 1 quote · 43 bookmarks · 15,379 views

  • @beffjezos

    Quotes Musk's $1,000 trillion reply as the answer to the question of SpaceX's total addressable market.

    423 likes · 33 reposts · 15 replies · 2 quotes · 38 bookmarks · 21,041 views

  • @beffjezos

    Quotes a boats-and-swimming analogy for AI in science and says superintelligence will open new "intellectual continents".

    336 likes · 23 reposts · 14 replies · 1 quote · 40 bookmarks · 15,112 views

  • @beffjezos

    Shares a video with the line that slowing technological progress is the greatest risk to humanity's persistence.

    309 likes · 42 reposts · 17 replies · 8 quotes · 33 bookmarks · 10,352 views

  • @PeterDiamandis

    Says solar met about three quarters of the growth in world electricity demand in 2025 and fossil generation did not rise.

    214 likes · 18 reposts · 12 replies · 1 quote · 11 bookmarks · 13,135 views

  • @agentsmart

    Shares a report that solar generation passed nuclear globally for the first time, contrasting slow nuclear growth with renewables and batteries.

    67 likes · 21 reposts · 7 replies · 1 quote · 2 bookmarks · 836 views

Engagement as returned by the X API on 2026-10-10 05:20 UTC

Read via the X API: recent posts from @beffjezos and @elonmusk, plus searches for e/acc, #eacc, effective accelerationism, kardashev, Dyson sphere, terafab, energy abundance, terawatt, petawatt, space solar, fusion and nuclear power, gigawatt and data center power. Kept on-topic posts (energy, compute, space, Kardashev, e/acc), skipped token promotion, off-topic posts and partisan political posts. Engagement is exactly what the API returned at fetch time.

X WATCH: 92 ANALYSED POSTS →