Insight · 409A & Complex Securities

When the last round isn’t valuation truth.

In 18 months, OpenAI was “valued” at $157 billion, $300 billion, $500 billion, and $852 billion, through a primary, a structured round, an employee tender, and a recapitalization. Backsolve still works, but the era when it could work casually is over.

Backsolve is one of the most useful tools in private-company equity valuation. When a financing round is recent, arm’s length, and representative of the security and market being valued, working backward from that observed price calibrates an option-pricing or hybrid model efficiently. The current AI mega-round and mega-tender environment has not broken that logic, but it has stress-tested it. The same enterprise can produce four different observed prices in eighteen months across different securities, different markets, and very different deal settings, and only one or two of them are likely to be the right anchor for any specific common-stock valuation date.

That is where the AICPA’s December 2025 working draft and the SEC staff’s Financial Reporting Manual converge. Transaction evidence matters, but only after the valuation specialist determines what market, what security, and what facts the price actually reflects. The era when a board could point to “the last round” and stop the analysis there has ended. The companies that have made that the most visible are the AI leaders.

OpenAI in eighteen months

Figure 1
One enterprise, four different prices, across different securities, markets, and stories.
Oct 2024
Primary · Convertible Notes
$157B
$6.6B raise. Convertible notes tied to a structural change, with negotiated protections if the change didn’t occur. Valuation embeds the contingency.
Mar 31, 2025
Primary · SoftBank-led
$300B
$40B announced. Large second tranche contingent on restructuring. Headline valuation, but partially conditional.
Oct 2025
Employee Secondary
$500B
~$6.6B of stock sold by current and former employees. Different security, different market, different seller motivation than the primary rounds above.
Oct 28, 2025
Recapitalization
Restructured
OpenAI Foundation 26%; Microsoft ~27%; remaining 47% held by current and former employees and investors. Cap-table architecture changes.
Mar 31, 2026
Primary · Strategic-led
$852B
$122B closed. Anchored by Amazon, NVIDIA, SoftBank, with continued Microsoft participation. Strategic-investor weight raises buyer-specific synergy questions.
Primary financing Employee tender Recapitalization event
The same company over eighteen months. Each observed price is real, and none of them is automatically the right anchor for a specific common-stock valuation date. Sources: company announcements; Reuters reporting on October 2025 employee secondary.

Anthropic at the other extreme

OpenAI shows price proliferation across security types. Anthropic shows a different stress: extreme operating velocity between transaction dates. On February 12, 2026, Anthropic announced a $30 billion Series G at a $380 billion post-money valuation, disclosing $14 billion in run-rate revenue, more than 500 customers spending over $1 million annualized, and over $2.5 billion of run-rate revenue from Claude Code alone. By April 2026, Anthropic announced run-rate revenue had surpassed $30 billion and the over-$1M-annualized customer count had doubled to more than 1,000.

A February price is not necessarily wrong, but it is not self-evidently current by April. Calibration that ignores the velocity between the round and the measurement date stops being calibration and becomes a snapshot stretched past its useful life.

Why the secondary market makes this universal

The AI names are extreme, but the dynamic is not. Carta’s data show 396 tender offers conducted on its platform in 2025, up 62% from 244 in 2024, with nearly 20% from Series E+ companies. Over the twelve months ending June 2025, Carta estimated VC secondary transaction value at $61.1 billion, above the combined value of VC-backed IPOs in the same period. PitchBook’s annual figure for 2025 puts U.S. venture secondary transaction value at roughly $106 billion.

396 +62%
Tender offers on Carta in 2025
$61.1B
Carta-tracked VC secondary value, July 2024–June 2025
$106B
PitchBook U.S. VC secondary transaction value, 2025

For any late-stage 409A engagement, that’s the new fact pattern: multiple observed prices of the same enterprise, across different securities and markets, often within months of the measurement date. None of those data points is junk, and none of them is automatically dispositive either.

Boards get into trouble when they treat the last round as a verdict rather than as evidence.

Five stress tests for backsolve

Figure 2
Before the price becomes the anchor, ask all five.
01
Primary & Secondary Spread
When primary and secondary prices diverge, which one matters for this measurement?
OpenAI’s March 2025 primary was $300B; the October 2025 employee secondary was $500B. The spread reflects operating progress, different rights, different motivations, and different markets, not a choice of which is “right.” Both are evidence, and weighting depends on relevance.
02
Structured Rounds
Did the headline price embed a contingency the model isn’t capturing?
Convertibles tied to corporate restructuring, second tranches contingent on milestones, and asymmetric downside protection all change what the price actually means. The AICPA draft says contingent transactions still require adjustment before they can support calibration.
03
Stale Round
How much did the company change between the transaction and the measurement date?
Anthropic’s February run-rate of $14B reached $30B by April. In a company compounding at that rate, a “recent” round can become stale faster than a board calendar. Calibration without a velocity adjustment is a snapshot, not calibration.
04
Strategic Investor Rounds
Does this price include something a generic market participant wouldn’t pay for?
The AICPA draft says calibration may not provide reliable evidence when there are no new investors, or when a round is led by a strategic with existing investors tagging along. Buyer-specific synergies should ordinarily be factored out of fair value.
05
Insider, Bridge & Compensatory Dynamics
Did the company organize, facilitate, or benefit from the transaction in a way that makes it more than a market signal?
Chapter 9 of the AICPA draft asks specifically: did the company benefit, did management negotiate, did the company set transaction parameters, did the price spread little versus a preferred round despite materially different securities, were buyers limited to current or former grantees? If yes, the trade may include a compensatory element under ASC 718, and it may be inappropriate to incorporate directly into common-stock fair value.
None of these tests rejects backsolve. Together they translate observed evidence into the price market participants would actually pay for the security being valued, on the date being measured.

What the SEC is asking

Recent comment-letter responses make the discipline visible. In 2025 correspondence, Sionna Therapeutics described its common-stock valuations as using OPM, a market-adjusted approach based on a recent arm’s-length transaction, and a hybrid PWERM/OPM with IPO and trade-sale scenarios. The company then had to reconcile a much higher IPO range to earlier fair-value conclusions by walking through scenario probabilities, DLOMs, and post-valuation milestones. In a March 2025 response, Jade Biosciences described hybrid valuations in which scenario probabilities shifted as a merger and financing became probable.

The staff is not banning backsolve. It is asking issuers to explain why last-round economics are, or are not, representative of current fair value for the shares being valued, which is the question every late-stage 409A board should be ready to answer in writing.

The questions reviewers will ask

Predictable reviewer questions, answered first
  • Why is this the right anchor security?
  • What rights did that investor receive that common stock does not?
  • Is this really the principal market?
  • What changed in the business since the round?
  • If this was a tender, who set the rules and who benefited?
  • Why was OPM chosen instead of PWERM, hybrid, or simulation?
  • If the company is preparing for an IPO, how would this conclusion stand up to SEC scrutiny?

The takeaway

In the AI mega-tender era, a transaction price is evidence rather than verdict. The better the round matches the subject security, the principal market, and the measurement date, the more persuasive the backsolve. The more the facts drift into structured terms, strategic motives, insider dynamics, selective secondary liquidity, or explosive operating change, the more the appraiser has to translate the price rather than repeat it. Far from skepticism for its own sake, that translation is what makes the work defensible, and it separates a 409A report that survives audit, board, and SEC review from one that becomes the start of a longer conversation.