AI chipmaker Cerebras, after withdrawing its上市 plan in 2025, resubmitted its IPO application to the SEC on April 17, 2026, aiming to list on Nasdaq under the ticker "CBRS." The company reported $510 million in revenue for 2025, a year-on-year increase of approximately 76%, with a net income of $87.9 million, compared to a net loss of $485 million in 2024. As of the end of 2025, the total value of unfulfilled performance obligations reached $24.6 billion. OpenAI previously provided Cerebras with a $1 billion loan and warrants, and the two companies have entered into cooperation agreements exceeding $20 billion. Morgan Stanley and Citigroup are serving as lead underwriters.

Canva and Anthropic have jointly launched Claude Design, powered by Claude Opus 4.7, enabling users to generate editable visual content aligned with brand guidelines through text descriptions. This release coincides with Canva AI 2.0, marking Canva's transformation from a design platform to an AI-first platform. AI 2.0 introduces conversational design, agent orchestration, and integrations with tools such as Slack and Gmail. The enterprise version supports automatic application of brand design systems. A research preview is currently available to Pro, Max, Team, and Enterprise users. Canva achieved $3.5 billion in revenue in 2025, with 265 million monthly active users and a valuation of $42 billion.

Anthropic CEO Dario Amodei met with the White House Chief of Staff on April 17 to discuss government access to the Mythos model. Mythos can identify and exploit thousands of zero-day vulnerabilities, successfully developing exploit programs on the first attempt in over 83% of cases. The model offers restricted access via Project Glasswing to around 40 vetted institutions. Previously, the company was labeled a national security supply chain risk after Amodei refused to remove safety restrictions for Pentagon use in autonomous weapons. Nonetheless, multiple government agencies continue seeking its defensive applications. This meeting may pave the way for a compromise allowing the Department of Defense to rescind the blacklist.

AI coding startup Cursor is in advanced discussions with investors to raise approximately $2 billion in funding, targeting a post-money valuation exceeding $50 billion. Existing investor Andreessen Horowitz will co-lead the round, with Nvidia also planning to participate and Thrive Capital in talks to join. If completed, Cursor would become one of the most valuable startups globally, reflecting strong market enthusiasm and capital support for AI coding tools.

Manycore (Kujiale) listed on the Hong Kong Stock Exchange on April 17, becoming the first among the 'Six Dragons of Hangzhou' to go public and hailed as the 'first global stock in spatial intelligence.' Priced at HK$7.62 per share, its public offering was oversubscribed 1,591 times. On its debut, the stock surged over 130%, with market capitalization surpassing HK$30.5 billion. The company maintains a database of 480 million 3D models and 500 million spatial scenes. It generated RMB 820 million in revenue in 2025 and achieved profitability for the first time. IDG Capital holds approximately 13% and is the largest institutional shareholder.

Factory announced on April 17 the completion of a $150 million funding round, reaching a $1.5 billion valuation, led by Khosla Ventures with participation from Sequoia Capital, Insight Partners, and Blackstone. Founded in 2023, the company focuses on developing an AI coding agent platform for enterprise engineering teams, supporting multiple foundational models including those from Anthropic and DeepSeek. It has established partnerships with enterprise clients such as Morgan Stanley, Accenture, and Palo Alto Networks. Keith Rabois has joined the board of directors.

Zoom announced a partnership with World, Sam Altman’s company, to launch a 'Human Verification' feature powered by World ID Deep Face technology. It uses triple verification—comparing Orb registration signatures, device facial scans, and meeting video frames—to display a 'Verified Human' badge. Meeting hosts can enable a verification waiting room. This move responds to rising deepfake video fraud; in Q1 2025, financial losses from deepfake-related scams exceeded $200 million, with enterprises averaging over $500,000 per incident.

Agibot unveiled the concept of the 'Year of Deployment' at APC2026, launching a full-stack embodied AI architecture called 'One Body, Three Minds' and seven comprehensive solutions, covering five hardware platforms including the humanoid robot A3, quadruped robot D2 Max, and high-dexterity robotic hand, along with eight foundational AI models. The company has set a five-year target of achieving RMB 100 billion in revenue. Its 10,000th general-purpose robot rolled off the production line at the end of March. Meanwhile, JD launched the industry’s first full-chain embodied AI data infrastructure, and Keenon Robotics opened the world’s first humanoid robot-staffed café in Shanghai.

Anthropic is transitioning its enterprise billing from a fixed per-seat subscription model to a token-based usage model, requiring customers to commit to minimum monthly spending thresholds. Under the new structure, access fees for Claude Code and Claude.ai platforms are $20 and $10 per month respectively, with actual usage billed separately at API rates. This change eliminates previous 10%-15% API discounts for large customers and is expected to increase overall costs for most enterprises. Existing customers must switch to the new model upon renewal. This reflects a broader industry shift toward pricing models tied to compute resource consumption.