Anthropic's IPO Filing - What Investors Need to Know

Summary

The presenter reviews Anthropic’s planned IPO, explains common IPO and lock-up risks, and assesses whether the company’s growth and financial position justify buying its shares. The central conclusion is to wait for Anthropic’s first public-company financial report rather than buy immediately after trading begins.

IPO mechanics and examples

An IPO makes a company’s shares publicly tradable. Before listing, private investors typically have less information. After listing, public companies must issue audited quarterly reports and disclose material risks.

The S-1 filing provides initial information, but investors may not see a company’s first public earnings report until roughly 90 days after the IPO. Early shareholders and employees are generally restricted from selling their shares during a lock-up period, often around six months. When restrictions lift, insider sales can increase the supply of shares and put pressure on the price.

The presenter cites these examples:

• Palantir (PLTR): Went public on September 30, 2020, at $10 per share and reached $39 four months later, a 290% gain. By the end of March, it had fallen 44% from its peak after a large share unlock. The presenter says about 80% of shares became eligible for sale on February 18.
• Arm (ARM): Relisted on September 14, 2023, at $56 per share. Its first-day gains quickly faded, and the shares later fell nearly 25% from the first day’s trading price. The presenter says the stock dropped by double digits around the expected lock-up expiry, but SoftBank did not sell. ARM later traded above $200 per share.
• Cerebras: Began trading the previous May at $350 per share and was around $110 at the time of the video, roughly 60% below its first-day price. Its lock-up was due to end November 9; the presenter had said he would wait until November to buy.

Anthropic IPO and share unlocks

Anthropic makes the Claude AI models. The video says its shares are expected to trade on Nasdaq under the ticker ANT, with a launch expected the week of November 16.

The IPO could raise about $100 billion at a reported valuation of $1.5 trillion–$2 trillion, potentially making it larger than the prior record IPO cited by the presenter. Private-market share prices cited were $5.61 in early 2025 and $58.90 the following May, described as roughly a tenfold gain in 15 months.

Anthropic’s S-1 reportedly lists 23 unlock dates. The presenter highlights the following:

• Two trading days after the first public earnings report, executives and certain employees may sell another 35% of their shares.
• Remaining shares unlock after the next report or 180 days on the exchange, whichever comes first.
• The seven co-founders may also sell up to 29% of their shares.

The presenter expects earnings reports and related unlocks to be important potential sources of selling pressure.

Business model and competitive position

Anthropic’s strategy is built around AI scaling: training larger models with more data and computing resources is expected to improve performance. The founders’ work on a 2020 paper helped establish these scaling concepts. The company says research breakthroughs—described as “computational multipliers”—can improve performance beyond what scaling alone would deliver.

The presenter describes Claude’s development in three broad stages:

1. Pre-training: Predicting the next token from large datasets of text and code. More data and compute are intended to improve performance.
2. Post-training: Turning the model into an agent that can plan and carry out tasks. Reinforcement learning rewards better outcomes and discourages weaker ones.
3. Inference: Customers use deployed models, generating token-based revenue and substantial serving-compute costs.

Anthropic also uses Constitutional AI, in which Claude checks responses against a set of principles. Claude is designed to run on Amazon Trainium, Google TPUs, Nvidia GPUs, and AMD GPUs, potentially allowing Anthropic to shift workloads toward better price-performance options.

The S-1 reportedly highlights Claude’s strength in enterprise work, especially programming. Risks include competition from OpenAI and Google, as well as free or lower-cost open-weight and open-source models. The video names Meta’s Llama and DeepMind’s Chinchilla as competitors.

Other risks identified include a plateau in scaling gains, expensive training runs that fail to pay off, customers switching models quickly, and AI agents requiring more compute than other products.

Financial performance cited

Figures below are as reported in the subtitles and attributed to Anthropic’s filings or reported results:

• Revenue rose from $41 million in 2023 to $386 million in 2024 and $4.6 billion in 2025.
• First-half 2026 revenue was $16.3 billion, reportedly 14 times the year-earlier period. Second-quarter revenue was $11.6 billion, implying an annualized pace of about $46 billion.
• The presenter says revenue grew 841% in 2024, 1,088% in 2025, and 1,368% year over year in Q2 2026.
• Customers paying more than $100,000 annually reportedly grew from nearly 1,000 to nearly 12,000 in 12 months. Nearly 2,000 customers paid more than $1 million annually. No individual customer reportedly accounts for more than 10% of revenue.
• About 80% of revenue comes from token usage, much of it not covered by long-term customer contracts.
• Reported gross margin was about 85% in the first half of 2026, versus about 40% in 2025, based on reported cost of revenue. Including model-training compute costs, the presenter estimates gross margin at about 45% in the first half of 2026, versus –60% in the prior year.
• In Q2, revenue reportedly grew 1,368% year over year, while computing costs rose 300%. The presenter says total computing costs fell from 251% of revenue early in the prior year to 48% in the latest quarter.
• Anthropic reported a GAAP net loss of $80.7 billion for the first half of 2026. The presenter says $72.4 billion was a non-cash expense, largely related to convertible notes whose value increased with the company’s valuation.
• The company reportedly generated $1 billion in operating profit in Q2, an operating margin of about 8%, and $3.9 billion in operating cash flow in the first half of 2026. Cash and short-term investments were $131.7 billion at the end of June.

Key investment concern: compute commitments

Anthropic reportedly signed about $600 billion in computing and data-center capacity commitments as of June—also cited as $587 billion—roughly 13 times its current annualized revenue. Some commitments reportedly run for 20 years, making them fixed obligations even if demand slows.

The reported order backlog was only $10.8 billion, less than a quarter of annualized revenue. The presenter sees this as a risk, while noting that token-based usage can vary and is not necessarily tied to long-term contracts.

Because AI agents may consume more computing resources, the presenter warns that margins could deteriorate if Anthropic cannot charge enough to cover the additional costs.

Investment approach and conclusion

The presenter’s framework for evaluating the IPO is to:

• Avoid assuming the initial IPO price is attractive, accounting for the limited public financial history and possible post-lock-up selling.
• Wait for the first public-company earnings report, expected early the following year.
• Track total computing costs relative to total revenue, including training costs, rather than relying only on reported cost of revenue or headline gross margin.
• Check whether compute costs continue falling as a share of revenue, particularly as AI-agent usage grows.
• Consider the timing of insider unlocks, especially the release of an additional 35% of certain insiders’ shares after the first report.

The presenter is positive about Anthropic’s revenue growth, enterprise adoption, customer diversification, improving compute efficiency, and cash position. The main caution is the scale and duration of its compute commitments compared with its relatively small reported backlog. His stated plan is to wait for the first earnings report and assess cost efficiency, rather than buy at the IPO.

Assets and companies mentioned

• Stocks and tickers: Anthropic (ANT, expected ticker); ARM; Palantir (PLTR); Nvidia; AMD; Meta; Alphabet/Google; Amazon; SoftBank; SpaceX; Cerebras (no ticker stated).
• Other technologies and models: Claude, OpenAI models, Meta’s Llama, DeepMind’s Chinchilla; Amazon Trainium, Google TPU, Nvidia GPUs, AMD GPUs.
• No explicit bond, ETF, commodity, or cryptocurrency investment is discussed.

Disclosures and sources

• The video includes a Wispr Flow / Wispr Flow Note Taker sponsorship.
• No formal “not financial advice” disclaimer appears in the subtitles.
• Presenter/source: Alex, from Ticker Symbol: YOU.