Base activated Cobalt on September 30. For issuers of its B20 tokens, the fork adds a way to schedule balance multipliers, seize balances with a record and combine transfer policies. None of that makes a tokenized asset a share in the company it tracks. It makes the rules enforced by the token more explicit, and the identity of the issuer more consequential.

The Cobalt upgrade specification records a September 30 mainnet activation, one week after Sepolia. Base’s public status page put the mainnet maintenance window at 18:00 UTC and marked it complete at 20:00 UTC. The fork adds B20 asset functions, transactions conditional on chain state, a registry for future upgrade scheduling in monitoring mode and an on-chain method for registering certain trusted-execution-environment prover signers. These are separate changes. The asset story begins with B20, the token format introduced with the earlier Beryl upgrade.

The fork went live, but its entire wish list did not

Two ideas that appeared in earlier Cobalt discussions are absent from the deployed scope. Payment of network fees in B20 tokens was removed from the fork’s list on September 29. Faster canonical 200 millisecond blocks belong to a proposed later Denim upgrade, not this activation. Native account abstraction has no scheduled mainnet gate here. Treating any of those as live Cobalt functions would confuse a roadmap with code an issuer or trader can use today. The distinction is especially important for institutions evaluating a token standard against current compliance requirements.

Base’s node release floor is v1.4.2 for mainnet according to the upgrade documentation. The preceding v1.4.1 included the timestamp but missed changes to validity transaction RPC forwarding; v1.4.0 does not contain the mainnet activation. A node that follows a fork without forwarding the new transaction type correctly can present a partial view of what users think is a uniform network. The code-level distinction is more instructive than a blanket statement that Cobalt is live.

The news hook is real, but it is not the whole thesis. B20’s earlier activation had already put issuer-managed assets on Base. Cobalt increases the actions the issuer can take and the policy decisions a transfer may face. The key question is who can invoke those functions, under what legal promise, and how a holder can check the result.

A B20 balance is a ledger entry with an issuer

A tokenized equity product can be represented as a balance on Base while rights to the underlying security sit with a broker, custodian or contractual issuer. The token standard cannot itself force a transfer agent to recognize the wallet holder as a shareholder. The link between on-chain balances and off-chain property rights comes from the product documents and the entities responsible for backing, redemption and corporate actions. A security token can be technically transferable and contractually restricted at the same time.

Coinbase’s tokenized-stock launch coverage describes a market in which backing arrangements and eligible users matter as much as trading interfaces. That context makes Cobalt’s additions more than developer conveniences. A policy can exclude an address, require a condition, or allow only a class of transfer. An administrative seizure can reassign a balance. A multiplier can change how balances display across accounts. Each function can support a lawful operational need and can also create dependence on an issuer’s judgment or administrative key security.

The B20 format needs to be examined at the token level. The mere fact that Base supports seizeWithMemo does not grant every B20 issuer a seizure right against every token, let alone every ERC-20 on Base. The B20 precompile reference says a token whose issuer has not configured the applicable policy slot has no seizure capability. An auditor has to inspect that token’s policy and authorized accounts. Two assets using the same standard can have sharply different holder rights.

This is the first control split to put on a whiteboard: the chain decides whether a transaction conforms to the deployed rules; the issuer decides which permitted administrative call to send; and the real-world asset provider is responsible for whether the token matches an enforceable claim. Cobalt changes the first two layers. It does not resolve the third. The same address may trade a token on-chain and still fail an off-chain eligibility test at redemption.

Seizure leaves a trace, but the reason is off-chain

Cobalt introduces seizeWithMemo, an issuer-authorized operation that moves tokens from a holder in one administrative step, superseding an earlier burnBlocked flow. The memo can leave a reason marker in the on-chain record. It does not prove the reason was legally sufficient. A smart contract can verify that the calling account has authority and that configured exemptions apply. It cannot decide whether a court order was valid, whether the issuer matched the correct defendant or whether a customer’s complaint should succeed.

The issuer operations guide describes the mechanics. A token might use seizure for a sanctions order, mistaken issuance, recovery under contractual terms or a corporate action. Each is a different justification. The holder should be able to find the administrator identity, the policy, the event and a dispute process in the product’s legal documents. If an issuer only says that tokenization is transparent, a reader should ask transparent about what: the transfer may be visible while the underlying decision remains opaque.

There is a subtle implementation detail. The documentation says the exemption scope changed name from SEIZE_HOLDER_POLICY to SEIZE_EXEMPT_POLICY, with a different selector. Code that hardcodes the old scope can fail to read or set the new one, even though older Beryl selectors otherwise continue. This is a genuine integration question for issuers and auditors, not a general claim that balances became newly seizable on September 30. Check the live token’s policy configuration and test the administrative call under the deployed fork.

The chain provides an evidence trail that conventional account corrections may not expose publicly. If an issuer moves 100 tokens from one wallet to another, observers can count 100 tokens and identify the transaction. They cannot infer a 100-share transfer in the issuer’s off-chain shareholder register without reconciliation. The strongest issuer case is that regulated assets need procedures for error correction and legal orders; tokenized stock volume on Base gives the practical context for why those procedures are now design choices rather than abstract debates. The trade-off is that a holder accepts an administrator with meaningful power.

The multiplier can change units without a matching deposit

A scheduled multiplier allows an issuer to define a future change in how a B20 asset’s unit balance is represented. Think of a stock split. If a holder’s displayed quantity moves from 10 units to 20 at a 2-for-1 ratio while the economic claim per unit halves, value need not change. The on-chain mechanism can coordinate the balance adjustment without asking each holder to sign. The issuer still has to implement the corresponding real-world corporate action and explain the conversion to brokers, custodians and price feeds.

The arithmetic is simple and the reconciliation is not. Suppose 1 million token units are outstanding and a 2-for-1 multiplier is scheduled. The new displayed units would be 2 million if the same multiplier applies across the relevant balances. That does not create 1 million additional underlying shares. A responsible issuer must show that total beneficial claims are unchanged and that the reference security’s own split took effect on matching terms. If token units double while a trading system keeps an old price-per-unit reference, a chart or collateral engine could misstate exposure by a factor of two.

The scheduling function improves coordination by naming the moment before it arrives. It also gives observers something to monitor: a pending update, its authorized signer and the post-change supply and holder balances. It does not guarantee every dependent system consumes the update on time. An exchange order book, oracle, lending vault and tax ledger can each use a different snapshot. A multiplier that is correctly executed on-chain can still create operational errors where integrations cache the old representation.

B20’s balance semantics also matter for historical data. An explorer showing the holder’s balance after the split may not explain how many units the holder held a day earlier or what each unit represented. Analysts should normalize quantities to the multiplier in force at each timestamp before claiming that deposits surged or supply inflated. A published event log gives a path to that normalization, but it is work someone has to do. Trading volume stated as raw tokens across the event is not comparable without an adjusted unit.

Combining policies exposes the eligibility decision

Union and Intersect are the two new composite policy types. Union permits an operation if an underlying policy accepts it under the configured logic; Intersect requires multiple underlying conditions to pass. The exact constituent policies and direction of authorization must be read from the token configuration. The useful analogy is a gate with alternative badges versus a gate requiring several badges. It is not a statement that every token must perform identity checks.

Imagine an asset whose issuer allows transfers to approved broker wallets or to a designated redemption contract. A Union policy can express alternatives. Another issuer may require that both the sender and receiver meet separate conditions, where an Intersect arrangement is more appropriate. If one condition is maintained off-chain through an authorized registry, the apparent on-chain transfer rule still depends on an organization updating that registry. A changed allowlist can change tradability without the holder moving a token.

Composite policies make it easier to describe a regulated asset in reusable modules. They can also make it harder for a holder to discover why a transfer failed if the interface reports only a generic revert. The B20 invariants and tests give developers a starting point, but a product still needs human-readable disclosure of which addresses can act, who updates lists and how errors are challenged. A permissioned token with an undocumented gate is not meaningfully transparent just because the gate is on a public chain.

The strongest opposing argument is practical. A tokenized security offered across jurisdictions cannot promise unrestricted transfer and also satisfy eligibility restrictions, court orders and corporate-action processing. Programmable controls can be more predictable than manual freezes in a broker database. That case holds when controls are narrowly delegated, auditable and tied to enforceable terms. The opposing risk is equally specific: one administrative key, policy registry or issuer interpretation can determine a user’s access. The Cobalt fork supplies primitives. Issuers supply governance.

Conditional transactions do not override issuer rules

Cobalt also introduces validity transactions: signed transactions paired with conditions on chain state, held until those conditions match. This is a general transaction feature, not an automatic compliance waiver. A user may want an order to execute only if a balance, price-related state or other predicate has a specified value. A transaction that becomes eligible still has to satisfy the token’s transfer policy at execution. If an issuer changed an allowlist in the meantime, the transaction can fail or remain ineligible depending on its conditions.

That interaction creates a valuable question for market structure. If a trader signs an order today that becomes valid tomorrow, who can change the state on which its execution depends? Some state comes from neutral contracts; some comes from an issuer-controlled policy. A conditional transaction can reduce one form of execution uncertainty while leaving the holder exposed to an administrator’s ability to update permissions. Integration documents should say which predicate was checked, when it was checked and what happens on expiry or cancellation.

Node software determines whether wallets and service providers see the new path reliably. The mainnet v1.4.2 minimum includes RPC behavior for forwarding validity submissions to a compatible sequencer ingress. A v1.4.1 node may follow the consensus fork but fail that submission route. For a user, the distinction appears as a confusing rejected transaction, not a discussion of release tags. For an institution, it calls for end-to-end testing against the exact node version and RPC provider used in production.

There is another boundary: Base’s sequencing and eventual settlement infrastructure. An issuer policy is enforced in execution when the transaction runs; conditional submission does not give a user a guarantee about when a sequencer includes an eligible transaction. Nor does a fast on-chain receipt by itself settle a legal dispute over the underlying stock. Cobalt improves expression and admission of transactions. It does not collapse ordering, legal ownership and redemption into one proof.

The paperwork determines the asset beneath the token

A holder evaluating a tokenized share should start outside the chain: who owns the reference security, where is it held, what claim does the token confer, and who owes the holder at redemption? If the product is a derivative or contractual claim on an issuer, the holder may not have the voting or insolvency rights of a direct shareholder. Cobalt does not change that classification. Its added controls can implement terms already in the agreement or give an issuer new technical capacity that requires updated disclosure.

Coinbase’s expanding tokenized-stock list illustrates the speed at which product menus can grow. A familiar stock ticker on an app is not a substitute for the issuer’s legal entity name and the asset-specific terms. Some products are available only to certain users or jurisdictions. Restrictions can be enforced at onboarding, at transfer, at redemption or at all three. If on-chain transfer is open but redemption is permissioned, the secondary buyer may end up with a token they cannot redeem directly.

A transparent policy disclosure would list each administrator role, the functions it can call, whether a multisignature is required, whether powers are time-locked and how emergency changes are announced. It would map each on-chain power to a contractual clause. It would show the reserve or custody verification process and how a holder can contest a seizure. The number of on-chain wallets holding a token cannot answer these questions. A token can spread across thousands of addresses while one issuer retains decisive authority over every redemption.

Cobalt also makes an old word, “ownership,” harder to use casually. One person can own the private key controlling a wallet. Another entity can control token issuance and administrative transfers. A custodian can hold the reference share. A broker can control access to the market. A court can assert authority over the claim. Those rights may be legally coherent, but their allocation must be explicit. The chain cannot rescue ambiguous product documents by making one part of the ledger public.

Measure deployment by configured assets, not by fork status

The fork’s activation is verifiable at a block and time. Adoption of its new B20 powers requires a different count: how many live asset contracts actually configure the new policies, how many schedule multipliers and how many invoke seizure? A zero count shortly after activation would not mean the fork failed. It would mean issuers had yet to use those optional functions. A large count would not prove the assets are fully backed or that the controls are well governed.

A reproducible measurement would inventory B20 assets, check policy selectors at the same block height, identify administrator addresses and classify observed calls after September 30. It would separately count attempts that revert and successful state changes. It would avoid assuming that an asset called “stock” has an underlying share merely because its metadata says so. This is a better adoption measure than transaction volume, which may reflect speculative trading in tokens whose legal structure differs widely.

The limit of the current record is that the fork specification describes a capacity, not a complete registry of active issuers and their terms. There is no universal Cobalt setting that determines all tokenized asset rights. Each issuer may configure functions differently, and a later update can change permissions. A node can correctly verify a transfer while the off-chain custodian’s statement remains late or disputed. A token can show an administrative move in public without telling the holder whether it was lawful.

The conclusion is concrete. Base now has more precise tools for issuers to control asset balances and eligibility. Holders gain a better chance to inspect those controls if issuers disclose them clearly. The meaningful test begins at each token: who can change the multiplier, who can seize, who can alter transfer policy and what legal claim survives if the issuer fails?

A holder can test three promises against one contract

The first promise is supply. A backing report might say that every token corresponds to one unit of an underlying asset held by a custodian. The holder can compare reported token supply at the report’s snapshot with the custodian’s stated position, adjusting for any multiplier then in force. The two numbers need the same timestamp and unit. A report of 1 million shares at yesterday’s close cannot be set beside a post-split supply of 2 million tokens today and called a deficit. Nor does a matching aggregate prove that every individual holder has the redemption right the marketing page implies.

The second promise is transfer. A product may advertise peer-to-peer settlement, then apply a policy that permits transfers only between registered intermediaries. Both statements can be true if the permitted peer set is narrow. A holder can inspect the token’s configured policies, submit a read-only simulation of a transfer between representative address types and compare the outcome with the published eligibility rules. That exercise should include a wallet eligible to hold, an ineligible wallet and the redemption destination. If results differ from the terms, the issuer should explain the mismatch before users trade.

The third promise is recourse. A user whose balance is moved by seizeWithMemo needs more than an event hash. The issuer should publish a case reference that protects private information while identifying the authority invoked, the applicable term, the date of notice and the channel for challenge. The holder can then compare the recorded movement with that account. A memo that says “compliance” without a procedure does little for someone contesting a mistaken identity or duplicated instruction. A token standard cannot compel a fair appeal, but its event trail can make the absence visible.

There is a fourth practical test for anyone using these tokens as collateral. A lending protocol may mark the asset to a market price and accept it as security for a loan. If the issuer can freeze or seize the collateral address, or alter the unit count through a multiplier, liquidation software needs to understand both events. A lender that prices an asset by ticker alone may miss a contract-level restriction on transferring it during liquidation. The borrower, meanwhile, may see a healthy price quote but be unable to move the pledged balance to repay. The relevant disclosure is whether the lending contract itself is exempt, who can change that exemption and what happens when the issuer revokes a user’s eligibility.

A custodian can answer some questions with an independent attestation, but an attestation has a scope. It might verify shares held in an omnibus account at a particular time without checking that token holders have a direct property interest. It might verify aggregate backing without checking whether a seizure changed the distribution among customers. A serious audit states the legal entity, the asset identifier, the snapshot time, the reconciliation method and the exclusions. The document should be refreshed after material issuance, redemption or corporate action. Readers should be able to compare successive snapshots, not merely admire a one-time badge on an app.

There is a failure case that the blockchain cannot settle: the issuer enters insolvency while the token continues to trade. The on-chain supply, policies and event logs may all be intact. The decisive question then is whether the underlying assets are segregated for holders, part of a custodian’s estate, or a general claim against the issuer. A smart contract with perfect enforcement of transfer restrictions does not pick a bankruptcy priority. This is why Cobalt’s administrative precision increases the urgency of reading product terms. It tells users exactly what the issuer can do with the token, while the documents must tell them what they can demand from the issuer.

An audit that tests all three promises would go beyond showing that code runs. It would reconcile the outstanding claims with assets held, the actual transfer gates with the published rulebook and administrative actions with a process outside the issuer’s own interface. The public chain supplies evidence for each test, but never the whole answer. Custody records and contract terms must be brought to the same date and unit. That is the work a tokenized asset requires after the celebratory fork announcement.

One more operational boundary deserves a public test. A token’s administrator might be a multisignature wallet with several signers, but a single company could appoint every signer. Publishing the threshold without naming the governing bodies does not show independent oversight. An issuer can disclose the threshold, key rotation procedure and emergency authority without revealing secrets. If it claims that holders can appeal a decision, it should identify the legal entity that reviews an appeal and the period in which it responds. These facts turn an on-chain permission into an accountable process.

A skeptical reader should also check whether policy changes emit events that data providers follow. If a wallet was allowed to transfer at noon and blocked at 12:01, the timing matters to a pending order, a lender’s margin calculation and a holder trying to redeem. A dashboard that updates once daily may make a real-time change look like a surprise seizure. Monitoring the contract directly can close that gap, but the product provider should still send notice to users whose rights change. Cobalt makes changes executable. Disclosure determines whether those changes are intelligible.

What to watch

• Configured policy counts: Public B20 contracts using Union, Intersect and seizure permissions after the September 30 fork.
• First scheduled multiplier: Its announced effective time, applied ratio and reconciliation with the off-chain corporate action.
• Administrative moves: Successful seizeWithMemo events, the authorizing role and an issuer explanation of each material case.
• RPC parity: Major Base node and RPC providers running at least v1.4.2 and accepting validity submissions consistently.
• Issuer disclosures: Product terms that map every on-chain administrator power to an enforceable right and appeal process.

FAQ

Disclaimer: This article is for information and educational purposes only and does not constitute financial or investment advice. Figures reflect regulatory filings and reporting available at the time of writing and change with each disclosure. Nothing here is a recommendation to buy, sell, or hold any security or asset. Always do your own research. Information is accurate as of October 1, 2026.

Stand With Crypto, the industry-aligned advocacy group launched by Coinbase in 2023 to push for what it describes as “clear, common-sense regulations,” has unveiled a slate of political endorsements ahead of the next congressional session. The organization says it is backing candidates across key Senate and House races that it believes could shape how US lawmakers handle digital-asset legislation.

Bloomberg Government News reported that Stand With Crypto plans to support Senate candidates running in 2026, including Ohio Republican Jon Husted, Iowa Republican Ashley Hinson, and New Hampshire Democrat Chris Pappas. The announcements come amid ongoing uncertainty around whether Democrats can regain control of either chamber of Congress in 2027, a factor that could determine what happens to major bills affecting crypto market structure.

Key takeaways

• Stand With Crypto is endorsing Senate candidates for 2026, including Jon Husted (Ohio), Ashley Hinson (Iowa), and Chris Pappas (New Hampshire), according to Bloomberg Government News.
• The group is also backing House candidates, naming Mariannette Miller-Meeks (Iowa) and Shomari Figures (Alabama), among others.
• Momentum for the CLARITY Act remains a political flashpoint after the US Senate failed to advance it through a cloture motion earlier this month.
• Crypto-linked political spending is already active, with Fairshake committing to major expenditures in the Ohio Senate race and spending more broadly by mid-year.

Why Stand With Crypto is making endorsements now

Stand With Crypto’s picks reflect a strategy of aligning electoral outcomes with regulatory priorities. The organization’s stated focus includes support for the Digital Asset Market Clarity (CLARITY) Act, a crypto market-structure bill that has moved unevenly through Congress.

Earlier coverage noted that the CLARITY measure failed to secure enough support in the Republican-controlled Senate after passing the House with bipartisan backing in 2025. In response to that failure, Stand With Crypto warned lawmakers who did not support the bill could face consequences, framing the moment as a high-stakes inflection point for future legislative direction.

The CLARITY Act and the Senate vote that set the tone

According to Stand With Crypto’s warning after the Senate’s Sept. 15 outcome, the inability to advance the bill would carry downstream effects for lawmakers. The group tied the issue directly to election dynamics, arguing that crypto policy is not just theoretical once votes are tallied.

The Senate’s earlier procedural setback matters because market-structure proposals typically require sustained coalition-building to progress from committee and chamber passage into law. With the CLARITY Act positioned as a top priority by the crypto policy community referenced in coverage, every procedural obstacle can shift both timing and bargaining power for follow-on legislation.

What remains uncertain is how quickly lawmakers will return to the same bill text—or whether they will pivot to alternative approaches—if the balance of power changes after the 2026 elections.

Endorsements and election control in 2027

Stand With Crypto’s announcement arrives against projections that Democrats may be poised to regain control of both the House and Senate in 2027. A shift in majority control could alter not only which bills move forward, but also how aggressively leaders prioritize crypto-focused regulation.

Bloomberg Government News reported the group is backing Senate candidates in 2026 who span party lines and states. The endorsements include Republican Jon Husted in Ohio, Republican Ashley Hinson in Iowa, and Democrat Chris Pappas in New Hampshire. On the House side, the same reporting says Stand With Crypto endorsed Republican Mariannette Miller-Meeks in Iowa and Democrat Shomari Figures in Alabama.

These endorsements also highlight a broader tension in Washington: while crypto policy advocates argue for regulatory clarity, procedural failures in one chamber can stall negotiations across the entire legislative pipeline. The election, in this sense, becomes a mechanism for shaping the legislative agenda rather than simply choosing representatives.

How crypto political spending is already shaping races

The backdrop for the endorsements includes substantial political activity from crypto-related groups. Cointelegraph previously reported that some candidates have already received significant funding ahead of the midterms, with money flowing through political action committees (PACs) and direct campaign contributions.

In the Ohio Senate race—where Stand With Crypto’s reported endorsement focuses on Jon Husted—Fairshake has been among the most prominent spenders. Cointelegraph reported that Fairshake pledged to initially spend $30 million supporting Husted against Democrat Sherrod Brown. The same reporting cited additional national spending, stating Fairshake spent $82 million on candidates across the country as of June, and that the organization has received backing from entities including Coinbase, Ripple Labs, and Andreessen Horowitz.

Stand With Crypto Executive Director Mason Lynaugh told Cointelegraph that there was “isn’t any evidence that crypto industry organizations are able to buy any assurances of who will win a race.” He attributed the organization’s endorsement of Husted to Husted’s vote in favor of the CLARITY Act, which he described as the industry’s “top policy priority,” rather than to broader assumptions about election influence.

Lynaugh also pointed to Brown’s earlier role in the Senate Banking Committee, characterizing it as influential in shaping crypto policy outcomes. At the same time, he said Brown’s more recent remarks about crypto’s role in the national economy were encouraging, while maintaining that endorsement decisions are grounded in voting records.

What to watch next

As Congress approaches another election cycle, investors, developers, and market participants should watch whether the endorsements translate into renewed momentum for market-structure legislation like CLARITY—especially if shifts in congressional control change which bills get prioritized and how quickly stalled proposals return to the floor.

Coinbase spent enough to make the top-ten in the overall category of securities and investment lobbying, according to OpenSecrets.org, above even Goldman Sachs Group Inc. and Andreessen Horowitz.

In U.S. law, lobbying must be legally disclosed through federal filings (though plenty of soft lobbying happens around the edges at parties and other events). CoinDesk’s review focused only on the first two quarters of the year — the most recent filings available, and nothing from 2025, which was also busy with crypto policy efforts.

Looking at the overall field of lobbying that didn’t explicitly flag the Clarity Act, four digital asset firms cleared the million-dollar threshold influencing crypto topics, including leader Coinbase, followed by $1.5 million from a16z, $1.4 million from Binance (all on outside lobbyists) and Crypto.com, which spent $1.2 million.

Industry companies and trade groups typically devoted the majority of their individual spending on their own lobbyist employees. When they sent checks to professional influence shops, that money scattered far and wide — to at least 42 distinct lobbying shops — though there were some firms that drew quite a bit more than others.

Checkmate Government Relations took in about $1.8 million in crypto-related money in the first six months of 2026 — most of that from Binance. The North Carolina operation is a relatively recent entrant into the lobbying field, though its client base is an epic list of corporate interests, including healthcare, technology, financial firms, tobacco companies and a major firearms manufacturer, and the firm is strongly associated with Republican interests and the administration of President Donald Trump.

Another of the industry’s favorites is Sternhell Group, run by Capitol Hill veteran Alex Sternhell. His firm took in $660,000 from digital assets names in those two quarters, and three of his four most lucrative lobbying clients came from crypto, according to the filings.

Four months ago, accredited investors in Anthropic’s Series H round were able to buy equity in the company at a $965 billion post-money valuation — a figure based on its 2025 book of financials.

The company’s new prospectus, leaked this week, however, shows the Claude AI owner preparing a November IPO seeking a $2 trillion valuation using those same 2025 numbers.

To justify its higher valuation, Anthropic has attempted to bridge the gap with unaudited, “run-rate” figures that annualize the month of May to incredible 12-month estimates.

It also disclosed Q2 financials yet used the month of May for annualization prospects.

Despite audited 2025 revenue of just $4.6 billion for an entire year of actual operations, Anthropic claims its “run-rate” as of May 2026 has become $47 billion — revenue growth of 1,000% that it somehow hopes to sustain for at least 12 additional months.

Runaway revenue and losses at Anthropic

Anthropic’s $4.6 billion revenue in 2025 incurred a $42 billion net loss due to the tremendous infrastructure costs of AI.

The company is trying to raise $100 billion at $2 trillion in what would be the largest IPO in history, topping SpaceX’s $85.7 billion haul at $1.77 trillion.

Nvidia has floated the idea of a $10 billion anchor stake in Anthropic’s IPO. Anthropic’s prior funding round closed on May 28, a $65 billion Series H with Altimeter, Dragoneer, Greenoaks, and Sequoia leading.