The Claude Fable ban barely changes Anthropic's IPO timing or valuation

To estimate the impact of the Fable situation (see my full forecast on how that will unfold), I did a full update of my Anthropic financial model through May 2027. This post is the diff between the old forecast and the new, as a way of measuring how much the Fable situation sets Anthropic back.

I think the change in Anthropic's overall situation is actually pretty small. I still forecast ARR near $93B in May 2027, a median IPO around December 2026, and a 90-day valuation near $1.1T. I updated my 90th percentile IPO date to be much later, and my 10th percentile valuation to be lower, but that's it. If the US government lifts the order, I think most of this reverses, though not all, because the market now knows the government can switch the flagship off.

Here are my updated views:

Where a range is shown, the first number is the median and the range in parentheses is the 80% confidence interval.

The rest of this posts justifies those updates, and puts the Fable situation in a broader context of how Anthropic the company is fairing according to my forecasts.

Anthropic brought its Mythos-class tier to general availability on June 9, 2026 as Fable 5 and Mythos 5, the first public models above the Opus line. On June 12, a Commerce Department export-control order, a Bureau of Industry and Security "Is Informed" letter under the Export Control Reform Act with foreign-national scope, forced Anthropic to disable both worldwide. They were still down on June 17, with the company and the administration in active talks (Anthropic). David Sacks framed it on June 13 as a problem of Anthropic's own making, saying the company "refused to fix" a jailbreak and calling the issue "easily resolved." Reporting traced the trigger to an Amazon-surfaced jailbreak, with Semafor citing suspicion of model distillation by Chinese labs.

The order hit two models, and I think the rest of the company kept running largely untouched. Claude Code, the fastest-growing line, runs on the Opus and Sonnet tiers the order left alone, so its forecast even rose, to $21B, on stronger enterprise adoption and a larger Cursor (now near $4B ARR) that still buys Claude API at scale. The capability frontier held too: Fable 5 posted 95.0% on SWE-bench Verified and 80.3% on SWE-bench Pro before it went dark, and those scores do not expire while the model sits suspended. So I would put the damage almost entirely in the tails of the financial forecasts and in the question of when the flagship comes back.

The clearest mark the ban left is on IPO timing, and I think it lands almost entirely on the tail rather than the median. The median barely moved, from December 20 to December 15, 2026. Before the ban, the late end of my 80% range sat at May 21, 2027, the edge of the forecast window; after it, that end runs to April 1, 2028, and the chance of pricing before May 21, 2027 falls from about 88% to about 78%.

I think the mechanism here is disclosure rather than demand. An active export-control order against the flagship is a material risk factor that has to go into the S-1, and a novel one, since it documents that the US government can switch off Anthropic's top product on short notice. SEC reviewers will work that section, and a single extra comment round can push pricing past the pre-holiday window into 2027, where the calendar and any unresolved order compound. I read the heavier mass at the far tail as the worlds where the standoff hardens and Anthropic waits rather than listing into an open question.

My median for the 90-day post-IPO market cap rose, from $1.05T to $1.09T, but I do not attribute that increase to the ban. It comes from the SpaceX comp: SpaceX priced its IPO at $1.77T and traded above $2T on its debut (see our SpaceX valuation), which re-rated the whole mega-cap-AI band upward. The ban shows up on the other side of the distribution, where the downside fell from $750B to $627B. An unresolved export overhang is the kind of risk public investors discount, on top of the gross-accounting question that was already there, so I think the ban widened the left tail even as an unrelated comp lifted the median. The distribution ends up wider and more left-skewed than the headline median move suggests.

The ban created a forecast that did not exist before, namely when the flagship comes back. I put restoration at a median of July 22, 2026, with an 80% range from late June to January 31, 2027, and I put the chance that a Mythos-class model is available and unsuspended on May 21, 2027 at 85%. I think that long tail reflects a real impasse: the administration has not named a clean technical fix, Anthropic is hardening its posture rather than folding, and the base rate for this kind of action argues against a quick resolution.

Pricing inherits the same uncertainty. Anthropic launched the Mythos-class tier at $50 per million output tokens, double the $25 Opus anchor, so absent the ban I would put the top GA model's price in May 2027 at a clean $50. The ban splits that forecast: about 85% of the mass sits at $50, where a Mythos-class model is available, and about 15% sits at the $25 Opus-only floor, the worlds where the premium tier never durably returns. The same split runs through capability, where the top GA model's SWE-bench Pro score is around 88% if a Mythos-class model is back and in the high-70s in the Opus-only tail.

The ban moved probability between scenarios more than it moved any single number. Before June 12 I had decoupling at 48%, convergence at 42%, and containment at 10%. The decoupling base case held at 48%, but I renamed it decoupling with friction, because the suspension is the template for the disruptions that case now expects. Convergence, the clean-détente bull case, fell to 30%. The mass went to the bear case, now entanglement, which rose from 10% to 22%: the world where federal friction institutionalizes, the premium tier stays under a restricted or permanent-control regime, and the IPO slips into late 2027 or 2028. I do not think the ban invented that world so much as showed it is more likely than I had it.

So what happens if Commerce lifts the order? I think most of the diff unwinds. The IPO's late tail tightens back toward a 2027 worst case as the risk factor moves from active to historical, the valuation downside narrows as the export overhang lifts, and the price forecast collapses toward a clean $50 with the $25 Opus-only tail mostly gone. Restoration resolves, near-term availability climbs toward certainty, and scenario mass flows from entanglement back to convergence.

It does not fully unwind, though, for two reasons. The first is that the precedent stands: June 12 proved the government can switch off the flagship, so an export-control risk factor stays in the S-1 even after a reversal, and public investors keep pricing some version of it. The second is that the likely shape of a reversal is a deal rather than a clean win, a US-person or compliance regime that institutionalizes a measure of government control rather than removing it. My rough read is that a reversal recovers most of the IPO-tail and downside damage, maybe two-thirds to three-quarters of it, but not all. The commercial core was never the thing at risk; what lingers is the risk factor the market learned in June.

For the full set of post-ban distributions, the federal-track detail, and the scenario write-ups, see Anthropic Revenue, Compute, and Valuation through May 2027.

Run this forecast yourself by

. Then ask Claude to refresh the numbers any time the news cycle moves.