Good morning and welcome back to another edition of The AI Rundown by Lightscape Partners.

• Intel warned that AI server demand exceeded its chip supply, leaving real data center revenue unrealized and spooking investors already focused on execution risk. The shortfall highlights how AI infrastructure booms can punish incumbents that lack manufacturing flexibility, even as demand for CPU and accelerator pairings accelerates faster than forecast.

• Meta signaled a reset in its AI trajectory after internal Superintelligence Labs delivered multiple major models in January, following a consolidation of research groups. The faster pace suggests Meta is prioritizing iteration speed and integration after criticism that recent releases lagged rivals, backing up its plan to spend aggressively on talent and infrastructure to stay competitive at the frontier.

• On the startup side, Humans& raised an eye-catching $480 million seed at a $4.48 billion valuation, one of the largest early rounds in AI history. Backed by Nvidia and Jeff Bezos and staffed by alumni from top labs, the deal shows how capital continues to concentrate around elite research teams, even as funding tightens for less differentiated AI startups.

Stay tuned as we explore these stories and their implications for the future of AI, technology, and innovation.

If you haven’t yet, please support the newsletter by subscribing!

Hardware

Intel says AI server demand outpaced its chip supply, forcing it to leave data center revenue on the table. Link.

• Intel said a spike in AI server orders caught it short on data center CPUs that pair with accelerators.

• Executives said factories are running full, but supply limits still constrain shipments to cloud customers upgrading for AI.

• The company forecast first quarter revenue and profit below expectations, amplifying investor concern about execution and capacity.

• Supply strain shows how AI infrastructure demand can punish vendors that lack flexible manufacturing and inventory buffers.

Alibaba prepares an IPO for its T-Head chip unit to help fund large AI infrastructure plans in China. Link.

• Alibaba is reportedly restructuring its T-Head chip arm ahead of a public listing with an employee ownership component.

• T-Head develops AI inference chips aimed at domestic customers seeking alternatives to Nvidia under United States export limits.

• An IPO would provide new capital as Alibaba commits tens of billions of dollars to expand AI and cloud infrastructure.

• Move highlights how Chinese firms are financing homegrown AI silicon to reduce reliance on imported high end accelerators.

Models

Meta says its new Superintelligence Labs delivered key AI models internally, signaling faster iteration after an organizational overhaul. Link.

• Meta CTO Andrew Bosworth said the team produced several major internal models in January after consolidating AI research groups.

• Reports indicate Meta is developing a next generation text model and an image video model targeted for early 2026.

• Progress aims to address criticism that Meta’s last major open model releases lagged rivals on capability and reliability.

• Faster internal delivery supports Meta’s plan to spend heavily on infrastructure and talent to compete in frontier AI.

Product Launches

Spotify rolls out Prompted Playlists, letting Premium users generate music playlists from natural language prompts. Link.

• Prompted Playlists lets Premium users in the US and Canada type a description and receive a custom playlist.

• The feature expands a prior test in New Zealand and builds on Spotify’s AI DJ concept for guided discovery.

• Spotify says prompts can request moods, situations, or introductions to an artist, lowering the effort needed to curate music.

• AI driven playlist creation is another retention bet as streaming services compete on personalization and user engagement.

Enterprise AI Applications

Gates Foundation and OpenAI launch a $50 million program to deploy AI tools in 1,000 African primary care clinics by 2028. Link.

• Horizon 1000 will equip 1,000 clinics with AI systems to assist health workers with diagnostics and administrative workload.

• The initiative starts with pilots in Rwanda and will expand through partnerships with African governments and local health programs.

• Backers said the goal is to improve care where clinician shortages are severe and resources for training are limited.

• Program tests whether practical AI deployments can deliver measurable health gains in low resource settings at scale.

Content Creation

YouTube plans a Shorts feature that lets creators generate videos using AI clones of their own likeness. Link.

• YouTube CEO Neal Mohan said creators will be able to produce Shorts featuring AI generated versions of themselves in 2026.

• The tools are expected to synthesize a creator’s voice and appearance, enabling faster production without being on camera.

• YouTube also previewed AI music capabilities, signaling a push to automate multiple steps in the creator workflow.

• Rollout will intensify debates on authenticity, consent, and platform safeguards as AI generated media becomes widely mainstream.

Startup Funding & Valuations

Humans& raises a $480 million seed round at a $4.48 billion valuation to build human centric AI collaboration tools. Link.

• Humans& was founded by alumni from OpenAI, Anthropic, Google DeepMind, and xAI, and raised $480 million at launch.

• TechCrunch reported the round valued the company at $4.48 billion, making it one of the largest seed financings in AI.

• Investors include Nvidia and Jeff Bezos, signaling strong demand for teams with frontier model experience and technical credibility.

• Funding wave shows capital still concentrates in elite research talent, even as many smaller AI startups struggle to differentiate.

Baseten raises $300 million at a $5 billion valuation as demand grows for AI inference and production deployment tooling. Link.

• Bloomberg reported Baseten raised $300 million and reached a $5 billion valuation, more than doubling in roughly six months.

• Baseten focuses on inference, helping companies run trained models in production with better performance and cost optimizations.

• The company raised $150 million in September at a $2.15 billion valuation, highlighting rapid repricing of AI infrastructure firms.

• Large rounds for inference tooling suggest buyers prioritize deployment economics as enterprises move from pilots to scaled AI use.

OpenEvidence raises $250 million at a $12 billion valuation to expand an AI medical search engine used by many US physicians. Link.

• OpenEvidence raised $250 million in a round led by Thrive Capital and DST Global, doubling valuation in three months.

• TechCrunch reported the company’s AI search product is used daily by about 40 percent of physicians in the United States.

• OpenEvidence says it limits training and retrieval to vetted sources and partners with medical publishers to improve accuracy.

• Fast adoption and high valuation reflect strong willingness to pay for clinical tools that reduce time and support decision making.

Inferact raises $150 million seed funding to commercialize inference software built by the creators of the vLLM open source engine. Link.

• Bloomberg reported Inferact raised $150 million in seed funding led by Andreessen Horowitz and Lightspeed Venture Partners.

• Founded by the creators of vLLM, Inferact plans tools that run large language models efficiently in production.

• Investors also include Sequoia, Altimeter, and others, indicating sustained market appetite for infrastructure that lowers inference costs.

• Deal highlights how open source performance breakthroughs can become venture backed companies when enterprise demand for deployment grows.

Regulation + Legal

TikTok reaches a deal to form a new US joint venture, shifting majority ownership to investors to avoid an American ban. Link.

• The agreement creates TikTok USDS, with ByteDance reducing its stake while a consortium of investors takes majority ownership.

• Reuters reported Oracle and Silver Lake are among the investors, aiming to keep data and operations under US controls.

• Both Washington and Beijing approved the compromise, ending years of legal fights over national security and influence risks.

• Outcome shows how platform governance disputes can force structural ownership changes to satisfy regulatory and geopolitical pressure.

Safety + Ethics

OpenAI introduces an age prediction system to apply stricter safety filters for teen users on ChatGPT. Link.

• OpenAI’s model estimates whether a user is under 18 using account signals, then applies tighter content restrictions automatically.

• Teens flagged by the system can verify age through a selfie process to regain access where appropriate and permitted.

• OpenAI said the goal is safer access rather than exclusion, with parental controls and limits on sensitive topics emphasized.

• Approach reflects growing pressure on AI platforms to manage youth safety without relying solely on self reported ages.

Meta pauses teen access to its AI characters on Instagram, Facebook, and WhatsApp while it builds a safer version. Link.

• Meta temporarily disabled teen access to AI character chatbots across its apps while developing a PG 13 compliant experience.

• Company said the relaunch will include parental controls and stricter topic boundaries focused on education, hobbies, and basic support.

• Move comes as Meta faces intensified scrutiny over teen safety and content moderation ahead of major legal and regulatory pressure.

• Pause signals that generative chat features may face rollout delays when safety controls lag or public trust erodes quickly.

OpenAI

OpenAI targets a late 2026 launch for its first consumer device, with reports pointing to an AI wearable that could be earbuds. Link.

• OpenAI confirmed plans to debut hardware in the second half of 2026 through its partnership with designer Jony Ive.

• Reports suggest the device, codenamed Sweet Pea, may run models locally on a custom 2 nanometer chip.

• TechCrunch reported ambitious shipment goals of 40 to 50 million units in the first year if production ramps smoothly.

• Hardware would give OpenAI direct distribution for assistant experiences, but it must compete with entrenched mobile ecosystems.

OpenAI launches OpenAI for Countries, offering governments help with AI data centers and deployments in public services. Link.

• OpenAI for Countries aims to support national AI projects including data center builds and AI integration for schools and hospitals.

• Reuters said 11 countries have already joined, with early deployments ranging from education tools to broader public sector pilots.

• Program positions OpenAI as a partner for state level AI infrastructure while competing with other labs and cloud providers.

• Government partnerships can accelerate adoption, but they raise questions about sovereignty, procurement transparency, and long term dependency.

Thank you for reading the AI Rundown by Lightscape Partners. Please send any questions, comments, or suggestions to [email protected].