AI Industry Daily Radar · July 10, 2026
Executive Summary
- OpenAI released GPT-5.6 — the Sol, Terra, and Luna model family — to the public on July 9, claiming new state-of-the-art results in coding agents, cybersecurity, and autonomous tool use, alongside a new "ultra" parallel-agent mode that runs up to 16 agents simultaneously.
- SK Hynix made its Nasdaq debut on July 10 after raising $26.5 billion in the largest foreign company IPO in U.S. history, pricing 177.9 million ADRs at $149 each on demand more than seven times oversubscribed.
- A Reuters exclusive revealed Meta will put its custom "Iris" AI inference chip into production in September, part of a plan to roughly double its computing capacity to 14 gigawatts by 2027.
- OpenAI also launched ChatGPT Work, a GPT-5.6-powered agent that autonomously produces finished documents, spreadsheets, presentations, and hosted websites — unifying Chat, Work, and Codex into a single desktop application.
- AI agent startup Lyzr used its own agent, SivaClaw, to run a $100 million Series B fundraise, fielding questions from over 130 investors without a founder ever flying to Sand Hill Road.
Top Stories
1. OpenAI Releases GPT-5.6 Sol, Terra, and Luna to the Public
Summary
OpenAI publicly launched the GPT-5.6 model family on July 9, 2026, after weeks of restricted access imposed by the U.S. government's voluntary model-review framework. The release includes three tiers — Sol (flagship), Terra (balanced), and Luna (most cost-efficient) — all available to developers and consumers simultaneously rather than through a staggered rollout.
GPT-5.6 Sol sets new state-of-the-art marks on several key benchmarks. On the Artificial Analysis Coding Agent Index v1.1, Sol scored 80, edging out Anthropic's Claude Fable 5 while using less than half the output tokens and roughly a third of the cost. On Terminal-Bench 2.1, Sol reached 88.8% (91.9% in "ultra" mode). The model also posted major gains in cybersecurity, scoring 73.5% on ExploitBench versus GPT-5.5's 47.9%.
The headline feature is "ultra" mode, which by default coordinates four agents in parallel — customizable up to 16 — to reduce wall-clock time on complex tasks. OpenAI also introduced "Programmatic Tool Calling," allowing the model to write and run lightweight programs that coordinate tools within the Responses API, reducing developer boilerplate. GPT-5.6 will be the default model across ChatGPT and the OpenAI API.
Source
https://openai.com/index/gpt-5-6/
2. SK Hynix Raises $26.5 Billion in Record Nasdaq Debut
Summary
SK Hynix, the world's second-largest memory chipmaker and the dominant supplier of high-bandwidth memory (HBM) critical to Nvidia's AI GPUs, began trading on Nasdaq on July 10 under the ticker SKHY. The company sold 177.9 million ADRs at $149 each, raising approximately $26.5 billion — the largest foreign-company IPO ever on a U.S. exchange.
Demand was extraordinary: the offering was more than seven times oversubscribed, forcing the company to close subscriptions early on July 8. Bank of America, Citigroup, Goldman Sachs, and J.P. Morgan led the deal. The listing removes what investors called an "accessibility discount" — many U.S. institutions previously could not easily hold Korean-listed shares. SK Hynix stock is up roughly 680% over the past 12 months, driven by what analysts describe as an AI-driven "memory supercycle."
Proceeds will fund new chip factories in South Korea and purchases of ASML's EUV lithography scanners. The stock is expected to join the Philadelphia Semiconductor Index, potentially unlocking significant passive investment flows. The debut underscores how AI infrastructure demand has become the single largest force in global capital markets.
Source
3. Meta's "Iris" AI Chip Enters Production as Compute Target Doubles
Summary
A Reuters exclusive based on an internal Meta memo revealed that the company will begin production of its custom AI inference chip, codenamed "Iris," in September 2026. Iris is part of Meta's four-generation MTIA (Meta Training and Inference Accelerators) project, designed in-house with assistance from Broadcom and manufactured by TSMC.
The memo also disclosed an aggressive infrastructure ramp: Meta plans to reach 7 gigawatts of computing capacity by end of 2026 (having added 1 GW in the first half) and 14 GW in 2027 — essentially doubling year over year. The company has secured multi-year supply agreements with Samsung for memory chips, Sandisk for flash storage, and Sumitomo Electric for fiber-optic equipment. Meta plans to release a new chip roughly every six months through 2027, a faster cadence than the industry norm.
The push reflects a strategic imperative to reduce dependence on Nvidia and AMD GPUs, which the memo acknowledged has been "a heavy lift" at Meta's scale. Meta expects to spend up to $145 billion on AI infrastructure in 2026 alone.
Source
4. OpenAI Launches ChatGPT Work, an Agent That Ships Finished Artifacts
Summary
Alongside GPT-5.6, OpenAI launched ChatGPT Work on July 9 — an agent inside ChatGPT powered by GPT-5.6 and built-in Codex technology that takes a high-level goal, gathers context from connected apps and files, and produces finished deliverables: documents, spreadsheets, presentations, reports, and even hosted websites.
The launch consolidates OpenAI's previously fragmented agentic offerings (Operator, deep research, ChatGPT Agent, Workspace Agents) into a single product. The standalone Codex app is merging into a unified ChatGPT desktop application with three modes: Chat, Work, and Codex. The Atlas browser is being sunsetted, with its agentic-browsing capabilities absorbed into ChatGPT. A new "Sites" feature (public beta) lets users build and share interactive web apps via URL.
ChatGPT Work rolled out immediately to Pro, Enterprise, and Edu users on web and mobile, with Plus and Business access following in days. OpenAI reports over 5 million weekly Codex users and 1,400+ connected plugins. Early testers like Zapier reported identifying "seven figures a month in pipeline." The move intensifies competition with Anthropic's Claude Cowork and Microsoft's Copilot Cowork.
Source
https://srnnews.com/openai-launches-chatgpt-work-adding-to-competition-for-professional-ai-tools/
5. GPT-Live: OpenAI's Full-Duplex Voice Models Enable Simultaneous Listening and Speaking
Summary
OpenAI released GPT-Live-1 and GPT-Live-1 mini on July 8, a new generation of voice models capable of full-duplex conversation — listening and speaking at the same time. GPT-Live-1 mini became the default voice mode in ChatGPT, replacing the previous Advanced Voice Mode that chained separate speech-to-text, LLM, and text-to-speech components.
The models enable natural interruption handling, live translation, and extended silent listening periods where the assistant absorbs context until explicitly summoned. ChatGPT Voice product lead Atty Eleti said he regularly conducts 30–40 minute conversations with the feature, and envisions voice becoming "the future primary interface to computing" for long-running agentic work. Over 150 million people already interact with ChatGPT via Voice and Dictation.
The release positions OpenAI against Apple's iOS 27 Siri upgrades, Amazon's conversational Alexa refresh, and startups like Sesame (founded by Oculus co-founder Brendan Iribe). Reports suggest OpenAI may launch AI-capable earbuds later this year, though no hardware was confirmed.
Source
6. Lyzr Lets Its Own AI Agent Run a $100 Million Fundraise
Summary
Lyzr, a three-year-old Jersey City startup that helps enterprises build AI agents, closed a $100 million Series B at a roughly $500 million valuation — and did so using its own agent, named SivaClaw, as the primary point of contact for investors. According to TechCrunch and Bloomberg, SivaClaw fielded questions from more than 130 investors, drafted investment memos, and tracked which pitch-deck slides backers lingered on.
The founders reportedly did not travel or conduct traditional in-person pitches; the entire process was managed remotely through the agent. Lyzr said it pulled in $400 million in investor interest against the $100 million it accepted. The raise is a striking proof-of-concept: the company's product literally ran its own fundraising process.
The story also illustrates the current AI funding frenzy, where capital is chasing AI deals so aggressively that founders with traction can raise nine figures without leaving their desks.
Source
7. Microsoft Launches $2.5 Billion "Frontier Company" for Enterprise AI Deployment
Summary
Microsoft announced Microsoft Frontier Company, a $2.5 billion consulting and engineering organization staffed with 6,000 industry and engineering experts, aimed at helping enterprises plan, deploy, and continuously improve AI systems. Led by Rodrigo Kede Lima and announced by Microsoft Commercial Business CEO Judson Althoff, the initiative represents a scaled-up version of the Forward Deployed Engineering (FDE) model.
Frontier Company engineers work on-site with customers to embed AI into existing products and business processes, iterate using real-world data, and deliver measurable ROI rather than proof-of-concept demos. Early adopters include London Stock Exchange Group, Land O'Lakes, Unilever, and Novo Nordisk. The group partners with Accenture, Capgemini, EY, KPMG, and PwC.
The move signals that Microsoft believes the biggest AI opportunity is no longer building better models — which are commoditizing — but helping organizations operationalize AI at scale. It follows AWS's $1 billion AI deployment commitment announced days earlier.
Source
Industry Trends
Trend 1: The Agent-as-Product Convergence
OpenAI's ChatGPT Work, Anthropic's Claude Cowork, and Microsoft's Copilot Cowork all launched within weeks of each other, each promising autonomous multi-step task execution that produces finished work rather than just answers. The frontier labs and their backers are racing to make the "agent that ships artifacts" the default computing interface. The consolidation of OpenAI's previously fragmented agentic tools into one product signals that the standalone-browser and separate-agent era is ending — the assistant itself is becoming the operating layer.
Trend 2: AI Infrastructure Spending Reaches Unprecedented Scale
SK Hynix's $26.5 billion listing, Meta's plan to double computing capacity to 14 GW, and Meta's $145 billion 2026 infrastructure budget collectively illustrate that AI capex has moved from "large" to "systemically significant." The SK Hynix IPO alone is the largest foreign offering in U.S. history, driven entirely by AI memory demand. Supply chains are straining — Morgan Stanley has dubbed the phenomenon "chipflation" — and every major player is vertically integrating into custom silicon to reduce GPU dependence.
Trend 3: Agents Begin Eating Enterprise Services
Microsoft's Frontier Company (6,000 experts, $2.5B), Lyzr's agent-run fundraise, and Anthropic's Claude Cowork plug-ins earlier this year all point to the same shift: AI agents are beginning to absorb work traditionally done by consulting firms, professional services, and even investment-banking processes. The fact that a startup could let its own agent close a nine-figure round is a tangible signal that the boundary between "AI tool" and "AI worker" is dissolving — with significant implications for the services economy.
Featured AI Products
ChatGPT Work
An agent inside ChatGPT powered by GPT-5.6 that takes a goal, gathers context from connected apps and files, and produces finished documents, spreadsheets, presentations, and hosted websites. It unifies Chat, Work, and Codex into a single desktop experience.
- Why it matters: It represents the most aggressive attempt yet to make an AI agent the primary interface for knowledge work, with 5M+ weekly Codex users as a built-in base.
- Official URL: https://openai.com/chatgpt
GPT-Live Voice Models
OpenAI's new full-duplex voice models (GPT-Live-1 and GPT-Live-1 mini) that listen and speak simultaneously, enabling natural interruption, live translation, and extended passive listening. GPT-Live-1 mini is now the default ChatGPT voice mode.
- Why it matters: With 150M+ people already using ChatGPT Voice, full-duplex capability could make voice the primary interface for complex agentic work.
- Official URL: https://openai.com/index/introducing-gpt-live/
Lyzr (SivaClaw)
An enterprise AI agent platform whose own agent, SivaClaw, autonomously managed a $100M Series B fundraise — fielding investor Q&A, drafting memos, and analyzing engagement data from over 130 investors.
- Why it matters: The most concrete proof-of-concept yet that AI agents can handle high-stakes, relationship-driven business processes end to end.
- Official URL: https://www.lyzr.ai/
Key Takeaways
- GPT-5.6 is OpenAI's most significant release of 2026, and the simultaneous launch of ChatGPT Work signals that the company's strategy has shifted from selling model access to selling finished work — a direct competitive move against Anthropic and Microsoft.
- SK Hynix's record IPO confirms that the capital markets now view AI memory and infrastructure as the most investable theme in technology, with HBM supply becoming a strategic bottleneck for the entire AI buildout.
- Meta's custom silicon push and 14 GW capacity target show that GPU dependency is the industry's biggest vulnerability — and every hyperscaler is now spending tens of billions to reduce it.
- The convergence of agent products (ChatGPT Work, Claude Cowork, Copilot Cowork) means 2026 is the year agents graduate from demos to default workflows.
- The Lyzr agent-run fundraise and Microsoft's Frontier Company both suggest the professional services and consulting industries are the next frontier for AI disruption.