AI News Digest, April 23: Physical AI Funding Hits Mega-Round Territory
Editor’s note: a Thursday digest from the Asanify newsroom, filed for HR and talent leaders watching how the AI money, the policy clocks, and the enterprise rollouts collide this week.
Physical AI funding just dragged the week’s headlines back into the venture column. Jeff Bezos’ Project Prometheus is closing in on a $10 billion raise at a $38 billion valuation, while Sarvam AI crossed into unicorn territory on the back of India’s largest pure-play AI round. Brussels set its next trilogue date for the Digital Omnibus, and CBIZ quietly turned 9,500 employees into Microsoft Copilot users. Here is the Thursday digest, with only the numbers and dates HR leaders should be able to quote in a board meeting.
Physical AI Funding Surges As Bezos Nears $10B For Project Prometheus
Physical AI funding has a new anchor tenant. Jeff Bezos’ stealth lab, code-named Project Prometheus, is in advanced talks to raise $10 billion at a $38 billion valuation, with BlackRock and JPMorgan among the investors leading the round, per Bloomberg and TFN, both citing the Financial Times. That is a 6x paper mark-up inside five months. Prometheus launched in November 2025 with $6.2 billion, co-led by Bezos and Vik Bajaj, and has scaled to more than 120 researchers pulled from OpenAI, xAI, Meta and DeepMind.
Why the name matters for HR leaders: Prometheus is not chasing another chatbot. It is training models on real-world experimental data, robotics interactions, and engineering workflows, targeting manufacturing, aerospace, and semiconductors. Bezos is separately reported to be raising up to $100 billion for a holding company that would acquire operating businesses and feed their data back into the lab. The physical AI funding thesis is that the next wave of enterprise value comes from models that can read a factory floor, not a PDF.
For HR operators, three things follow from this round. One, the talent market for robotics, reinforcement learning, and industrial ML researchers just tightened again, and counter-offers above $750K total comp will become normal at the senior end. Two, workforce-planning questions are moving from “will AI automate white-collar tasks” to “which blue-collar and field roles get re-skilled first,” which is a different policy and training conversation. Three, if your CEO starts asking about a physical AI funding strategy for your own product line, the honest answer is that frontier-scale capital and a two-year hiring sprint are table stakes. For most mid-market HR teams, the near-term play is integrating with the physical AI tools your vendors ship, not building them.
Bezos has not confirmed the round publicly. The close is expected within weeks.
Sarvam AI Closes India’s Largest Pure-Play AI Round
India’s sovereign AI story got its first unicorn this week. Sarvam AI is finalising a Series B of $300 million to $350 million at a $1.5 billion post-money valuation, with Bessemer Venture Partners leading and Nvidia, Amazon, Prosperity7 and HCLTech joining as strategic investors, according to Bloomberg and Outlook Business. Glade Brook Capital, a backer of xAI and Perplexity, is in talks to join. If the round closes as reported, it becomes the largest pure-play AI funding round in India’s history.
Sarvam’s edge is Sarvam-105B, a 105-billion parameter model tuned for 10+ Indian languages under the IndiaAI Mission’s sovereign-model track. For HR teams running operations in India, the practical signal is a native-language HRIS assistant is about to stop being a roadmap item at Indian SaaS vendors and start being a demo. Vernacular onboarding, payslip chat, and grievance triage in Tamil, Marathi and Bengali now have a domestic foundation model behind them.
EU AI Act Trilogue Heads For An April 28 Showdown
Brussels has the next EU AI Act trilogue pencilled in for April 28 as negotiators try to close the Digital Omnibus package before the August 2, 2026 obligation for high-risk systems bites. The Commission’s December 2025 proposal offered the industry hard compliance dates of December 2, 2027 for Annex III high-risk systems and August 2, 2028 for Annex I, a roughly 12-month slip that civil-society groups have pushed back on. Parliament’s IMCO committee is scheduled to vote on its position on May 13.
For HR teams, the key line item is unchanged: automated screening, scoring, and emotion inference for hiring or performance decisions remain classified as high-risk. Even if the slip holds, HR stacks that currently rely on AI resume screeners, video-interview scoring, or engagement-sentiment tooling should assume documentation, human-oversight logs, and data-quality evidence will be audit-ready by Q3 2027 at the latest. Waiting for the grace period to start is not a plan.
CBIZ Turns 9,500 Staff Into Microsoft Copilot Users
CBIZ, the US professional-services firm, announced on April 21 that it is rolling Microsoft Foundry, M365 Copilot, and Copilot Studio out to its entire 9,500-person workforce. The early result the firm disclosed is a 40% reduction in candidate-screening time inside its talent function, per Microsoft’s customer story.
The CBIZ case is worth reading closely because it is not a pilot. Nine thousand five hundred named seats, company-wide, on a single productivity-AI stack. For HR leaders building the business case internally, the proof point has shifted. In 2025, boards asked for AI pilots. In 2026, the comparable firms are asking why their own deployments are not yet population-scale. If your Copilot program is still gated to a “champions” cohort six months in, plan for that question at your next quarterly review.
Quick Hits
Gemini Enterprise gets a partner war chest. Google launched Gemini Enterprise, a unified agentic work platform, alongside a $750 million fund to seed Workspace-integrated AI agent startups. The target is enterprise workflow automation where Microsoft currently leads. HR implication: expect aggressive partner-led pricing on recruitment, onboarding, and learning agents that plug into Gmail and Drive over the next two quarters.
DeepMind raises the safety bar. Google DeepMind published v3 of its Frontier Safety Framework this week, adding Tracked Capability Levels and a new Critical Capability Level for harmful manipulation risk. It is the first time a frontier lab has put persuasion and manipulation capability on the same formal ladder as cyber and CBRN risk. HR teams fielding AI-written performance reviews and engagement nudges should note this. The category is now under active red-teaming at Google.
What This Means For HR This Week
Three calls for Monday morning. First, re-price your AI engineer comp bands against the Prometheus and Sarvam data points. The market is moving even if your last benchmark is ninety days old. Second, ask your ATS vendor for an EU AI Act high-risk compliance roadmap with a named owner and a Q3 2027 deadline, not a vague 2028. Third, if you run India operations, add Indic-language HR chat to your 2026 RFP criteria. The models to support it will ship this year.
For HR leaders who want a running view of which AI tools actually move hiring and engagement metrics, Asanify tracks production deployments inside our HRMS platform and publishes what works. Our field notes on AI agents in HR workflows, AI in recruitment, and the AI tools HR teams use in 2026 dig deeper into the deployments that actually ship.
FAQ On Physical AI Funding And HR
What is physical AI funding and why is it rising in April 2026?
Physical AI funding refers to venture capital flowing into companies that train AI models on real-world sensor, robotics, and industrial data, rather than text alone. The April 2026 spike is driven by Project Prometheus closing in on $10 billion at a $38 billion valuation, and several frontier labs signalling that the next margin gains will come from models that can operate in factories and logistics, not just chat.
How does the EU AI Act Digital Omnibus affect HR technology vendors?
The proposed Digital Omnibus pushes Annex III high-risk compliance to December 2, 2027 and Annex I to August 2, 2028, but the August 2, 2026 deadline for general-purpose AI obligations stays in force. HR tools that do automated screening, scoring, or emotion inference remain classified high-risk, and buyers should ask vendors for documentation, human-oversight logs, and data-quality evidence as standard.
What should HR leaders do about Microsoft Copilot deployments after the CBIZ rollout?
Move past the pilot phase. CBIZ’s deployment to 9,500 staff with a 40% reduction in screening time is a signal that the reference benchmark is now population-scale. HR leaders should define a 12-month path from champions cohort to full-staff rollout, with clear productivity metrics per function, or expect boards to ask why not.
Not to be considered as tax, legal, financial or HR advice. Regulations change over time so please consult a lawyer, accountant or Labour Law expert for specific guidance.
