Amazon’s Anthropic investment could reach $272 billion at IPO | Magica
Amazon’s Anthropic investment could be worth $272 billion at IPO
Editorial Team
••📖8 min read
Amazon’s shares and convertible notes could be worth $271.8 billion at the top of Anthropic’s proposed IPO range. Existing accounting gains, a 12% ownership cap and Anthropic’s computing commitments shape what that paper value means.
Amazon’s shares and convertible notes could be worth $271.8 billion at Anthropic’s proposed maximum IPO price of $78.
Amazon already recorded a combined $238.7 billion for those investments at September 30; restrictions limit immediate conversion and sale.
Anthropic’s $110 billion minimum spending agreement with Amazon gives the cloud supplier another financial stake in the AI developer’s growth.
Amazon’s $18 billion investment in Anthropic could be valued at nearly $272 billion if the AI developer prices its proposed initial public offering at the top of its range, according to calculations based on the registration filing. That would be about 15.1 times the money invested—a prospective gross value, before any sale or taxes.
Anthropic is the AI research and software company founded in 2021 that develops Claude, a family of models and applications used for coding and knowledge work. Claude’s commercial launch was in the first quarter of 2023. Its November 10 preliminary prospectus outlines an 825 million-share offering at $70 to $78 each, including shares sold by existing stockholders. It has applied to trade on Nasdaq under “ANT.” Pricing is still pending.
Amazon’s exposure extends beyond the securities. Its Amazon Web Services cloud division, AWS, supplies the computing infrastructure Anthropic needs to build and run Claude. The IPO brings attention to both the appreciation of Amazon’s investment and the commercial obligations linking the companies.
Amazon’s invested amount versus the prospective gross value of its shares and convertible notes at $78 a share. The conditional valuation represents neither realized proceeds nor an additional accounting gain. Source: Reuters via StreetInsider.
The estimate includes securities Amazon cannot immediately sell
The $18 billion invested comprises $8 billion in convertible notes bought in 2023–25 and $10 billion in preferred-stock purchases in the second quarter of 2026, according to Amazon’s quarterly filing.
JetBrains IDEs 2026.3 bundle Air for running coding agents and reviewing their changes, while the standalone Air app is in maintenance mode. Existing provider subscriptions remain usable, but billing, security controls and execution options vary by integration.
Amazon recorded $92.5 billion for its Anthropic nonvoting preferred stock and an estimated fair value of $146.2 billion for its convertible notes at September 30. Together, that was $238.7 billion before the IPO.
The prospective $271.8 billion value is $33.1 billion above those combined recorded amounts. That difference compares private-security accounting measurements at September 30 with a hypothetical valuation at the proposed IPO price. It does not establish the additional accounting gain Amazon would recognize when the offering closes.
Conversion also has limits. Anthropic’s prospectus lists $6.203 billion in outstanding Amazon notes that, without an ownership cap, could convert into about 1.84 billion nonvoting Class N shares. Amazon has elected to retain its 12% cap for the offering. The cap uses a capitalization measure that includes shares underlying equity awards and other convertible securities.
The principal that cannot convert within the cap remains in non-interest-bearing rollover notes. Further conversions can occur at subsequent quarter ends, subject to the cap, which Amazon may waive. The final offering size and capitalization will determine the immediate conversion. Amazon also expects a customary IPO lock-up, followed by continuing securities-law restrictions on sales.
Economic exposure does not confer equivalent voting power. Anthropic is a Delaware public benefit corporation, required to balance shareholder interests with responsible AI development for humanity’s long-term benefit and the interests of people materially affected by its conduct. Its Long-Term Benefit Trust, intended to be an independent body of financially disinterested trustees, generally has authority to elect and remove a majority of authorized board members. A separate share class to be held by an entity owned by the co-founders will carry 50.1% of the general stockholder vote. Amazon’s shares are nonvoting.
Financing and computing supply reinforce each other
The relationship was already established well before the IPO. In its November 2024 partnership announcement, Anthropic named AWS its primary cloud and training partner and said a new $4 billion investment would bring Amazon’s total investment to $8 billion. It also described engineers working with AWS’s Annapurna Labs on Trainium, Amazon’s custom AI chips, and the software needed to use them.
Amazon said in its April 2026 expansion announcement that more than 100,000 customers ran Claude models on AWS. It said Anthropic was already training and deploying Claude on Project Rainier, a cluster with nearly half a million Trainium2 chips. The expanded agreement envisaged access to up to five gigawatts of capacity; that figure describes planned access, rather than capacity already delivered.
Anthropic’s prospectus discloses a commitment to spend at least $110 billion with Amazon from May 2026 through April 2036, regardless of actual usage. A shortfall generally requires payment of the difference, subject to adjustments and exceptions. Anthropic cannot cancel except for Amazon’s uncured material breach.
A further $15 billion Amazon financing facility becomes available in $5 billion increments as specified computing-capacity milestones are reached. Those future funds are separate from the $18 billion already invested. Amazon’s quarterly filing also records contractual obligations related to the performance of its chips: the arrangement ties financing to infrastructure delivery as well as tying Anthropic to spending.
Long commitments can address an immediate constraint. Anthropic says customer demand has exceeded available computing capacity. It says $5 billion of compute delivered under short-term commitments in the third quarter cost significantly more than comparable capacity under multi-year commitments, reducing revenue per dollar of compute from the preceding quarter. The contracts reserve future supply, while exposing Anthropic to payments even if usage disappoints.
Amazon is a major partner among competing suppliers
Google, part of Alphabet and another cloud supplier, could have an interest worth $139 billion at $78 a share, according to the same filing-based calculations. Together, the two companies’ prospective interests would total roughly $411 billion.
The investment bases differ. Google’s $33.3 billion investment figure includes a planned $10 billion purchase in the private placement accompanying the IPO. Chipmaker Nvidia has also agreed to buy $10 billion of shares in that placement. The prospectus says the $26 billion placement across seven technology companies would close immediately after the public offering, subject to conditions, and issue nonvoting Class N shares at the IPO price.
Anthropic also has substantial minimum spending agreements with other cloud providers. The prospectus gives these amounts and periods; they cover different agreements and durations and do not establish comparable unit prices:
Cloud provider
Minimum committed spending
Agreement period
Amazon
$110 billion
May 2026–April 2036
Google
Approximately $121.3 billion
April 2026–July 2033
Microsoft
$47.2 billion
November 2026–June 2033
Claude is distributed through AWS, Google Cloud and Microsoft Azure. Anthropic says it maintains direct customer contracts, sets pricing and pays cloud distribution partners a fee. It trains and serves Claude across Amazon Trainium, Google TPUs, Nvidia GPUs and AMD GPUs, giving it alternatives to a single chip supplier. Amazon, meanwhile, separately invested $50 billion in rival AI developer OpenAI, according to its quarterly filing.
These overlapping roles raise competition questions. A January 2025 staff study by the Federal Trade Commission, the U.S. competition and consumer-protection agency, examined Amazon–Anthropic alongside Alphabet–Anthropic and Microsoft–OpenAI. It identified commitments to spend substantial portions of cloud partners’ investments on their services and flagged potential switching costs and access to sensitive business information. The study used information available through January 2025; it was not a finding on the proposed IPO or the subsequent contracts.
Company-reported revenue and compute and infrastructure expenses for the nine months ended September 30. Expenses include research, model development and internal workloads as well as customer-serving computing; the gap is not operating profit. Source: Anthropic SEC registration statement.
The next test is delivery and demand after pricing
Final pricing and capitalization will determine how much of Amazon’s investment converts immediately. Lock-up restrictions and the eventual trading price will shape its opportunities to realize that value.
The operating business supporting the valuation has expanded sharply. Anthropic’s financial disclosures show $33.242 billion in revenue for the nine months ended September 30, 2026, against $2.545 billion a year earlier—about 13.1 times as much. Its operating loss narrowed to $1.260 billion from $5.224 billion over the same periods; the earlier loss included a $1.5 billion litigation settlement charge. Compute and infrastructure expenses rose to $18.808 billion from $4.381 billion. Those expenses cover research, model development, customer-serving and internal workloads, so they are not simply the cost of serving paying customers.
Its $104.039 billion net loss included $94.324 billion in fair-value changes on convertible notes, derivative liabilities and other financing instruments. That loss is not a measure of operating cash consumption. Anthropic generated $6.6 billion in operating cash flow during the nine months and held $95.6 billion in cash and cash equivalents plus $36.3 billion in short-term investments at September 30, before IPO proceeds.
The prospectus puts aggregate committed spending under purchase commitments and lease agreements, including arrangements entered into after September 30, at approximately $900.9 billion, substantially all for computing and infrastructure. The commitments span many years, with substantially all spending after 2030 expected by 2045. About $70.7 billion was expected to be paid in the 12 months following September 30, 2026, including $4.2 billion tied to arrangements with contingent conditions or termination rights before commencement.
The evidence to watch after pricing is whether suppliers deliver capacity on schedule and whether customer usage supports the spending as it comes due. Those results will bear on both the value of Amazon’s securities and the cloud revenue promised by the partnership.
Nvidia chief Jensen Huang argued in May 2025 that US chip restrictions were strengthening Chinese rivals. Filings show a substantial commercial stake, but lost sales and market share alone cannot establish whether the controls failed their security purpose.
Reported demand has Anthropic’s bankers considering a price above its filed $70–$78 range. The combined financing could exceed $100 billion, but existing cash, uneven operating results and long-term computing commitments complicate the valuation test.
The British AI chip designer has secured private and UK government backing before its planned 2027 commercial launch, putting a multibillion-dollar valuation on its promise of faster, cheaper inference.
Zhang Yiming reportedly told staff that ByteDance needs years to challenge leading US AI labs, even as the TikTok parent prepares a personal agent for release within weeks. Its consumer reach and video strengths have yet to translate into frontier language models.
Helsing has unveiled a military intelligence platform promising European control over data, models and operations. Germany's evaluation of roughly 30 European systems makes procurement and demonstrated performance the next tests.
Public safety leaders warn that synthetic voices and background sounds make hoax emergency calls harder to assess. A voluntary survey exposes preparedness gaps, but does not establish a national AI-swatting rate or measure growth over time.
Fortinet's November security disclosures include AI-assisted findings, but some affected software branches still await fixes. Earlier testing put token costs at about $540 per true positive, excluding the broader work of reviewing findings and securing customer deployments.
Tamil Nadu told the Madras High Court that accounts containing about 2.22 million rupees had been frozen after its library-procurement order, while a criminal case was still being prepared. The dispute tests safeguards ordered in 2024; AI involvement remains unconfirmed.
Unions withdrew objections to Google’s proposed $10 million purchase of Spirit Airlines data after securing employee-record exclusions and independent privacy review. The agreement follows earlier passenger-data protections; court approval and verification remain ahead.
Baird cut Nvidia to Neutral and its price target to $200 from $500. Its case combines slower forecast AI spending growth with China-linked inventory exposure and uncertainty over founder and CEO Jensen Huang’s tenure.
ByteDance is reportedly accelerating plans to combine its Spell project with Doubao. Existing travel bookings reveal a transaction-fee business, while the phone assistant's history shows how access to other apps can constrain execution.
The Office for Students wants more AI-free assessment and better AI skills teaching, backed by existing quality rules. Its research raises doubts about unsupervised coursework, but universities say assessment reform is already under way.
Finland has recognised Tesla’s provisional Dutch approval after changes to branding and speed controls. Drivers remain responsible, and the national decision does not commit Finland to backing EU-wide authorisation.
Grok Bot can connect to internal MCP servers through Tailscale without exposing them to the public internet. User-based authorization preserves network controls, but administrators must configure how narrowly the assistant can reach into their systems.
Daylight’s Goldman Sachs Alternatives-led Series B backs security agents supported by human engineers. The young company reports rapid growth, but has not disclosed the economics behind its automation promise.
Primordial Labs raised $18 million to widen deployment of Anura, its military command software. Existing contracts and demonstrations give it a foothold, but integration counts do not establish sustained use or lower operating costs.
Northwestern Polytechnical University bought two Nvidia H200 processors for 874,700 yuan and awarded five contracts for rented Nvidia computing, procurement records show. Separate US restrictions on the defense-linked institution remained in place despite January’s conditional opening for H200 exports to China.
The North Dakota Republican says allegations involving a suspected Chinese intelligence operative make Huang vulnerable to coercion. Nvidia’s board is backing its founder while independent directors investigate, with investors facing both security and leadership risks.
Scott Bessent says Washington and Beijing are establishing a channel for serious AI incidents, following their September agreement. Its usefulness depends on unresolved reporting procedures and participation by the companies that detect model failures.