This week’s Alibaba AI share placing set the tone for a seven-day stretch in which the industry’s headline story was not a product launch or a model upgrade, but the sheer scale of money and legal exposure now circling the AI industry.
Alibaba’s Alibaba AI Share Placing: The HKD Numbers Behind the Headlines
The snippet figure of £7.5 billion is a currency-conversion approximation. Alibaba’s SEC filing gives the authoritative figures: the company placed 710 million new shares at HK$112.70 each, raising gross proceeds of HK$80.0 billion and net proceeds of HK$79.7 billion, with the placement closing on 26 August 2026.
The placing expanded Alibaba’s total outstanding ordinary shares by approximately 3.7%, to 19.885 billion shares. That is a meaningful dilution for existing holders, and investors will want to see the “full stack” AI build-out (new models, chips, and computing infrastructure) justify it quickly.
The timing is pointed. Alibaba’s quarterly profits reportedly plunged 75% on rising AI infrastructure spend, and the company was blacklisted by the Pentagon in June alongside other Chinese firms deemed a US national security risk. Raising HK$80 billion in Hong Kong, right after those headlines, is a statement as much as a capital raise: Alibaba intends to compete, and it is not waiting for Washington’s blessing.
Amazon’s Twitch Problem Is Bigger Than One Lawsuit
Not one but two proposed class actions have landed against Amazon and Twitch over the use of creator content to train AI models. The first, filed on 20 August 2026 in the US District Court for the Northern District of California, names Warren Pandiscia, a Connecticut creator who has streamed on Twitch for ten years and has more than 900 followers, according to FindLaw. A second proposed class action, reported by Bloomberg Law, names Eden Gordley and alleges violations of California’s right of publicity laws, characterising creator output as “recorded human performance” encompassing appearance, speech, mannerisms, and audience personas.
The scale of the platform these suits concern is substantial. Courthouse News Service reports that the 37-page Pandiscia complaint alleges Amazon had “an overwhelming incentive to acquire training data on an unprecedented scale” because its AI products are commercialised. Per the same reporting, Twitch had roughly 240 million active monthly users by 2026, with between 26 and 30 million people visiting the platform each day to view content from 3.2 to 6.9 million creators each month.
The opt-out that Amazon announced in August, after reportedly harvesting the content first, is precisely the kind of consent architecture that plaintiffs’ lawyers love to pick apart. Amazon and Twitch have not commented on either proceeding.
Anthropic’s $2 Trillion Question
Investor chatter around an Anthropic IPO worth more than $2 trillion would be easy to dismiss as the usual pre-listing hype, except the revenue trajectory makes the number at least internally coherent. The snippet put annualised revenues at $47 billion in May 2026. By the end of July 2026, that run rate had reportedly exceeded $65 billion, according to GraniteShares citing Reuters reporting. The $2 trillion figure is itself built on projected 2028 revenues of $190 billion to $200 billion, per the same analysis.
Anthropic confirmed it confidentially submitted a draft registration statement on Form S-1 to the SEC on 1 June 2026, according to the Anthropic newsroom. The number of shares and the offer price remain unset. A potential October listing is in the frame, but Anthropic could still walk away if conditions shift. The $2 trillion figure reflects what investors are modelling, not what the company has guided.
For context, SpaceX, which framed its own public offering partly as an AI infrastructure play, filed its Form S-1 with the SEC on 20 May 2026, with Goldman Sachs, Morgan Stanley, BofA Securities, Citigroup, and J.P. Morgan among the lead underwriters, per the SpaceX SEC filing. Two potential mega-listings. Two companies leaning hard on the “AI infrastructure” framing. Public markets will decide how much of that framing they are willing to price in.
The Other Stories Worth Watching
The Meta child-safety trial, which commenced on 18 August after 29 US states filed suit, is the sleeper issue for brands. A ruling that forces changes to infinite scroll and like counts would reshape the social media environments that fashion and beauty advertisers have built their upper-funnel strategies around. Worth keeping very close tabs on.
Meanwhile, the Phia affiliate attribution controversy is becoming a useful case study in what happens when AI shopping apps scale faster than their governance structures. The app is now appointing a head of compliance and reviewing every transaction. Whether that is sufficient to rebuild retailer confidence, and what it signals for how brands should vet their AI commerce partners, is the more durable question here.
The Anthropic October listing window is the next hard date to watch: if market conditions hold, it could set a public-market benchmark for frontier AI valuations that every subsequent IPO, and every brand partnership negotiation with these companies, will be measured against.
