One-Line Summary

Alibaba is behind the viral HappyHorse video AI, Meta locks in billions more compute, and OpenAI juggles record funding with a Mac app security scare — the video and infrastructure maps are shifting fast.

Big Tech

Meta commits to spending additional $21 billion with CoreWeave

Meta agrees to spend an additional $21 billion with GPU cloud provider CoreWeave, a move framed as part of the ongoing race for AI compute capacity. While CNBC highlights the scale, it underscores how hyperscalers now pre-book specialized infrastructure rather than relying solely on their own data centers. ¹

Reports add that this contract focuses on AI inference rather than training and could include early deployments of Nvidia’s next-gen Vera Rubin platform. Some coverage pegs the broader CoreWeave relationship at roughly $35 billion and notes CoreWeave plans to raise $4.25 billion in new debt to expand capacity. For teams, this signals more stable, lower-latency access to AI features across Meta’s apps. ²

Beyond CNBC, other outlets echo the $21 billion figure but with limited detail; the throughline is clear: Meta is diversifying compute vendors to ensure headroom for AI features. That redundancy matters if you build marketing or community on Instagram or WhatsApp: more backend capacity typically translates into faster feature rollouts and steadier performance. ³

OpenAI identifies security issue involving third-party tool, says user data was not accessed

OpenAI says a compromised update to the popular Axios JavaScript library briefly entered a GitHub Actions workflow used to sign macOS apps like ChatGPT Desktop, Codex, Codex-cli, and Atlas. The firm found no evidence of user data access, IP compromise, or altered software, but it’s revoking trust for older macOS builds and requiring updates. ⁴

OpenAI ties the incident to a broader supply-chain attack reportedly linked by some to North Korean actors and says the signing certificate in the impacted workflow was likely not exfiltrated. Effective May 8, older macOS app versions may stop working. If your team uses the desktop app, plan to update across all devices this week. ⁵

Axios notes that attackers briefly published two infected Axios updates on March 31 before detection. OpenAI says it hasn’t seen impact beyond macOS, but the episode is a reminder: even legit open-source libraries can be a risk. If you manage devices, this is akin to rotating keys and re-issuing badges after a building vendor incident. ⁶

OpenAI raises $122 billion to accelerate the next phase of AI | OpenAI

OpenAI says it closes a round with $122 billion in committed capital at a post-money valuation of $852 billion, citing deep participation from strategic partners including Amazon, NVIDIA, SoftBank, and continued involvement from Microsoft. The company frames compute as the core moat and claims it now generates $2B in revenue per month. ⁷

External coverage portrays intensifying rivalry with Anthropic, which some reports say is growing fast in enterprise accounts. Commentary highlights that both firms are racing to secure massive compute commitments, with OpenAI arguing its larger capacity will convert into product advantage as AI demand surges. For non-technical teams, this tug-of-war often shows up as faster model refreshes and new agentic features in everyday tools. ⁸

A separate report asserts Anthropic’s annualized revenue exceeds $30B, edging past OpenAI’s self-reported $24B annualized figure; methodologies differ, but the takeaway is enterprise usage is now the growth engine. For you, that means vendors will prioritize workplace-grade reliability, governance, and coding assistants — areas where budgets are flowing today. ⁹

Industry & Biz

Alibaba just revealed it’s behind a viral AI video model dominating leaderboards

Alibaba confirms it created HappyHorse-1.0, the previously anonymous video model that shot to the top of Artificial Analysis’ blind-test rankings for text-to-video and image-to-video this week. The reveal came via a new X account tied to Alibaba’s ATH AI Innovation Unit; Alibaba verified the post as authentic. Shares in Hong Kong closed up 2.12% after the news. ¹⁰

Bloomberg reports Happy Horse hit the top spot on debut and remains in beta. For context, Alibaba has been expanding its AI suite around its Qwen model family and integrating AI into e-commerce, ads, and entertainment — video generation could slot directly into product listings, shoppable videos, and ad creatives. ¹¹

The timing is notable: OpenAI recently discontinued its Sora video app and platform, while ByteDance paused Seedance 2.0 amid copyright disputes. That leaves room for new leaders. For marketers and PMs, think of HappyHorse as a potential “video copilot” for rapid product demos or concept reels — even if you can’t use it yet, this is a signal to prepare prompts, brand guidelines, and approval flows for text-to-video. ¹⁰

Analysis pieces suggest Alibaba’s anonymous launch let the model prove itself before brand scrutiny — like submitting a dish to a blind tasting before revealing the chef. If Alibaba follows its usual playbook, expect tight integration into commerce and media surfaces where conversion lifts matter most. ¹²

South Africa unveils draft AI policy, proposes new institutions and incentives

South Africa releases a draft national AI policy for public comment, proposing a National AI Commission, an AI Ethics Board, and an AI Regulatory Authority to coordinate policy, oversee standards, and monitor compliance. The plan prioritizes investment in data centers and networks and flags concerns about reliance on foreign infrastructure. Comments are invited by June 10. ¹³

Local coverage says the framework rests on six pillars, including capacity and talent, inclusive growth and jobs, responsible governance, ethical and inclusive AI, cultural preservation and international integration, and human-centered deployment. For companies operating in or with South Africa, this signals upcoming guidance on procurement, data handling, and incentives. ¹⁴

In parallel, Africa-focused outlets point to startup activity around efficient AI — like model compression for edge devices — as governments push both innovation and oversight. That context matters if you run operations across bandwidth-constrained regions: policies plus local tooling could make on-device AI more viable for customer support and offline analytics. ¹⁵

What This Means for You

• Video gen is going mainstream in commerce and ads. Alibaba’s HappyHorse leapfrogs benchmarks just as other big names stall, which usually precedes product embedding. If your team runs PDPs, TikTok Shop, or shoppable reels, start drafting prompt templates and brand-safe guardrails now — the bottleneck moves from “can we?” to “what’s on-brand and legally safe?” ¹⁰

• Compute is the new supply chain. Meta’s multibillion-dollar CoreWeave spend is like booking cargo space before holiday season — it ensures AI features can be served reliably at peak. For users, that often means snappier AI assistants in Instagram, WhatsApp, and Ads Manager — and fewer slowdowns when new features drop. ¹

• Enterprise AI is where budgets flow. OpenAI’s massive raise and the reported surge at Anthropic tell you buyer priorities: governance, reliability, and coding/co-pilot tools that ship measurable productivity. If you manage budgets, expect more vendor proposals emphasizing audit trails, workspace controls, and ROI on agentic workflows. ⁷

• Security ops need “supply-chain” muscle memory. OpenAI’s Mac certificate scare wasn’t a data breach, but it’s a clear reminder that third-party libraries are risk gateways. Treat AI desktop apps like any business-critical software: centralize distribution, pin versions, and schedule forced updates. ⁴

Action Items

1. Update ChatGPT macOS apps across your team: OpenAI says older versions may stop working after May 8 — open the app and install the latest update on every Mac you or your team uses. ⁵
2. Collect brand-safe text-to-video prompts: Draft 10 prompts and a 1-page brand/usage guideline for product demos or ad concepts so you can test any new video model quickly when it becomes available. ¹⁰
3. Watch benchmark samples to refine creative briefs: Browse Artificial Analysis’ video leaderboard clips and note what prompts yield strong motion and fidelity; translate those into your next campaign brief. ¹⁰
4. Enable centralized desktop app management: If you’re the ops/IT-adjacent person, standardize AI app installs via an MDM or shared checklist to avoid supply-chain risks from ad-hoc downloads. ⁶