Bryan Christmas

Working on something new

[Justin Schroeder @jpschroeder]

Fable isn't the first.

In 1999 the department of defense blocked exports of the PowerMac G4 for crossing the 1 gigaflop threshold.

Steve Jobs turned it into an ad.

[Bryan Christmas @bchristmas]

@holdyourwinners @ErikVoorhees @DavidSacks Would mean more if Amazon wasn’t the 2nd largest beneficiary of this (1st is OpenAI, massive Trump donor via Brockman). Regulatory complexity means more need for Bedrock/GovCloud, and giving OpenAI room to catch up just happens to look like Amazon commoditizing their complement

[Bryan Christmas @bchristmas]

@__tinygrad__ @wolftivy @Scobleizer @DarioAmodei @AnthropicAI I don’t think the cyber aspect alone would have ended up just undiscovered and un reported on for very long. Some blowback on one of these releases was inevitable, if not this one then 3 months from now, and with pitchforks if they had somehow been negligent

[Bryan Christmas @bchristmas]

@marlytokenman @scaling01 Sources close to the company say Dario could be heard screaming at staff, slamming his MacBook - “we’re ruined!!”

[Bryan Christmas @bchristmas]

@antoniogm @larryfink_jr He was writing about doing his startup, give the man some time to marinate

[Bryan Christmas @bchristmas]

@SemiAnalysis_ This isn’t the UN though, startups are already at war all the time and it’s the reason we have any access to such great models at all. Blocking distillation is fair play, and a lot of the safety stuff is real now

[Bryan Christmas @bchristmas]

@ethanmckanna @Waymo Happened to me recently, not even a bug that’s really how it works

[Bryan Christmas @bchristmas]

@ClementDelangue Great take, but the rub is where is the new balance for the cached tools… HF is using cool stuff in the first place, but most companies are paying big $ for saas plus a team just to “customize” it and get locked in… saas might be going from 3 TV channels to YouTube

[Bryan Christmas @bchristmas]

@agupta @nico_laqua @garrytan They’re also just not aligned or motivated to reply to everyone. I’m pretty sure no one is giving those people performance based comp plans because they know the apts are going to get leased regardless, but you can find a good one for less than $6k around dogpatch eventually

[Bryan Christmas @bchristmas]

@pitdesi @paularambles Even Google Maps team is TPU squeezed enough to run Flash-lite on lower Manhattan, tough times out there

[Bryan Christmas @bchristmas]

@benhylak yea probably ends up more like youtube than imovie

[Bryan Christmas @bchristmas]

@Teslarati Probably at a yellow, they should have it chill on that

[Bryan Christmas @bchristmas]

@benthompson Its cool, and it makes a lot of sense to not try to appeal to the same buyers with it for now. Trying to please both segments with one thing is how they’d end up as a Chinese company’s new sub brand

[Jason Crawford @jasoncrawford]

Most of our problems are caused by modernity, because modernity already solved most of the ones that came before.

[Bryan Christmas @bchristmas]

@Anand051098 @iruletheworldmo @chetaslua Yea looks like maybe it comes down to not being given access to tools when set to standard

[Bryan Christmas @bchristmas]

@GergelyOrosz This rule unfortunately applies to any Google product. Sad too, it really seemed like something was shifting for a bit, but lately… not so sure. Sergey is back to enjoying his life, I think is what happened

[Bryan Christmas @bchristmas]

@Jaselton @adcock_brett Worth questioning, but the training data almost certainly includes a few junk bits with people adjusting their headsets

[Jukan @jukan05]

Why did xAI hand over a 220,000-GPU cluster to Anthropic?

The technical backdrop to xAI's decision to hand Colossus 1 over to Anthropic in its entirety is more interesting than it appears. xAI deployed more than 220,000 NVIDIA GPUs at its Colossus 1 data center in Memphis. Of these, roughly 150,000 are estimated to be H100s, 50,000 H200s, and 20,000 GB200s. In other words, three different generations of silicon are mixed together inside a single cluster — a "heterogeneous architecture."

For distributed training, however, this configuration is close to a disaster, according to engineers familiar with the setup. In distributed training, 100,000 GPUs must finish a single step simultaneously before the cluster can advance to the next one. Even if the GB200s finish their computation first, the remaining 99,999 chips have to wait for the slower H100s — or for any GPU that has hit a stack-related snag — to catch up. This is known as the straggler effect. The 11% GPU utilization rate (MFU: the share of theoretical FLOPs actually realized) at xAI recently reported by The Information can be read as the numerical fallout of this problem. It stands in stark contrast to the 40%-plus MFU figures achieved by Meta and Google.

The problem runs deeper still. As discussed earlier, NVIDIA's NCCL has traditionally been optimized for a ring topology. It works beautifully at the 1,000–10,000 GPU scale, but once you push into the 100,000-unit range, the latency of data traversing the ring once around becomes punishingly long. GPUs need to churn through computations rapidly to keep MFU high, but while they sit waiting endlessly for data to arrive over the network fabric, more than half of the silicon falls into idle. Google sidestepped this bottleneck with its own custom topology (Google's OCS: Apollo/Palomar), but xAI, by my read, has not yet reached that stage.

Layer Blackwell's (GB200) "power smoothing" issue on top, and the picture comes into focus. According to Zeeshan Patel, formerly in charge of multimodal pre-training at xAI, Blackwell GPUs draw power so aggressively that the chip itself includes a hardware feature for smoothing power delivery. xAI's existing software stack, however, was optimized for Hopper and does not understand the characteristics of the new hardware; when it imposes irregular loads on the chip, the silicon physically destructs — literally melts. That means the modeling stack must be rewritten from scratch, which in turn means scaling is far harder than most of us imagine.

Pulling all of this together points to a single conclusion. xAI judged that training frontier models on Colossus 1 simply was not efficient enough to be worthwhile. It therefore moved its own training workloads wholesale onto Colossus 2, built as a 100% Blackwell homogeneous cluster. Colossus 1, on the other hand — whose mixed architecture is far less crippling for inference, which parallelizes more forgivingly — was leased in its entirety to an Anthropic that desperately needed inference capacity.

Many observers point to what looks like a contradiction: Elon Musk poured enormous capital into building Colossus, only to hand the core asset over to a direct competitor in Anthropic. Others read it as xAI capitulating because it is a "middling frontier lab." But these are surface-level reads.

Look at the numbers and a different picture emerges. xAI today holds roughly 550,000+ GPUs in total (on an H100-equivalent performance basis), and Colossus 1 (220,000 units) accounts for only about 40% of the total available capacity. Colossus 2 — built entirely on Blackwell — is already operational and continuing to expand. Elon kept the all-Blackwell homogeneous cluster (Colossus 2) for himself and leased out the older, mixed-generation Colossus 1. In other words, he handed the pain of rewriting the stack — the MFU-11% debacle — to Anthropic, while keeping his own focus on training the next generation of models.

The real point, then, is this. Elon's objective appears to be positioning ahead of the SpaceXAI IPO at a $1.75 trillion valuation, currently floated for as early as June. The narrative SpaceXAI now needs is that xAI — long the "sore finger" — is not merely a research lab burning cash, but a business with a "neo-cloud" model in the mold of AWS, capable of leasing surplus assets at high yields.

From a cost-of-capital perspective, an "AGI cash incinerator" is far less attractive to investors than a "data-center landlord generating cash."

As noted above, the most important detail of the Colossus 1 lease is that it is for inference, not training. Unlike training, inference requires far less tightly synchronized inter-GPU communication. Even when the chips are heterogeneous, the workload parcels out cleanly across them in parallel. The straggler effect — the chief weakness of a mixed cluster — is essentially neutralized for inference workloads.

Furthermore, with Anthropic occupying all 220,000 GPUs as a single tenant, the network-switch jitter (unanticipated latency) that arises under multi-tenancy disappears. The two sides' technical weaknesses end up complementing each other almost exactly.

One insight follows. As a training cluster mixing H100/H200/GB200, Colossus 1 was an asset that could only deliver an MFU of 11%. The moment it was handed over to a single inference customer, however, that asset transformed into a cash-flow asset rented out at roughly $2.60 per GPU-hour (a weighted average of the lease rates across GPU types). For xAI, what was a "cluster from hell" for training has become a "golden goose" minting $5–6 billion in annual revenue when redeployed for inference. Elon's genius, I would argue, lies not in the model but in this asset-rotation structure.

The weight of that $6 billion becomes clearer when set against xAI's income statement. Annualizing xAI's 1Q26 net loss yields roughly $6 billion in losses per year. The $5–6 billion in annual revenue generated by leasing Colossus 1 to Anthropic, in other words, almost perfectly hedges xAI's loss figure. This single deal effectively pulls xAI to break-even.

Heading into the SpaceXAI IPO, this functions as a core line of financial defense. From a cost-of-capital standpoint, if the image shifts from "research lab burning cash" to "infrastructure tollgate stably printing $6 billion a year," the entire tone of the offering can change.

(May 8, 2026, Mirae Asset Securities)

[Jukan @jukan05]

Elon’s SpaceX has submitted plans for its first fab, and the initial investment alone is reportedly $55 billion. If fully expanded, total spending could reach as much as $119 billion.

…Are they serious?

[Bryan Christmas @bchristmas]

@Nowooski It’s like the entire Bay Area has only this one slop architecture option for new developments. Most difficult approvals and yet somehow they end up looking like some kinda Temu Lego theme park