“Compute is worthless if you can’t power it.” With @cvpayne, I break down the I/O Fund’s early call on Bloom $BE at $16—and another company well positioned to meet the scale of power AI will require over the next decade.

Computers, whether quantum or classical, depend on algorithms. Quantum computing, however, needs a larger toolkit of standardized operations. @BrookhavenLab scientists and partners developed a quantum version of an important classical operation:

Computer-use agent experience = capability × robustness × efficiency. As frontier models climb higher on CUA benchmarks, efficiency becomes increasingly attrative, as highlighted by the recent release of #Jev#. With many intermediate steps in everyday computer-use tasks being “local,” fast System-1 models like #CUA-S1# and #Jev# can make agent execution much more efficient. Cool exploration by @francedot and @ddupont808 at @trycua!

compute situation is crazy i have an opencode session doing work with trainium and it had to make a reservation for capacity on saturday so it scheduled a task to wake itself up then so it can continue working

Computer scientist Ken Shirriff programmed an Apollo Guidance Computer to mine bitcoin. This was the onboard computer developed to navigate NASA’s Apollo spacecraft and control their engines during missions to the Moon. It weighed about 70 pounds and had roughly 4 KB of RAM. This particular unit never flew in space. Its owner had bought it from a scrapyard in the 1970s after NASA discarded it. Shirriff wrote custom code to make the 1960s machine perform SHA-256, the calculation used in Bitcoin mining. One calculation took 5.15 seconds, implying approximately 10.3 seconds for Bitcoin’s double hash. At the network difficulty in 2019, he estimated that finding a new block would take, on average, roughly a million times the age of the universe. His demonstration used a block that had already been mined. The computer reproduced the calculation; and never did earn bitcoin.

Compute infra feels a lot like the early crypto days. From bare metal to hardware financing, data center construction, chips, cooling, power procurement, grid interconnection, and state by state energy permitting. Then networking, Gpu/tpu utilization, workload scheduling, and inference optimization. And finally compute markets via onchain, trading infra, financing, pricing oracles, and verification that the compute you bought actually gets delivered. Hair on fire problems in almost every part of the stack. Pretty exciting times.

Compute had its moment, then memory had its supercycle. Marvell CEO Matt Murphy tells @PatrickMoorhead and @danielnewmanUV the next constraint for AI is connectivity, moving data between chips, racks, and data centers at AI scale. $MRVL

Compute futures are no longer just a concept. The next question is: how do you actually use them? Our new practitioner’s guide uses a 50-H100 GPU example to explain hedging, daily settlement and basis risk—step by step.