Seven developments in AI capital markets, developer tool pricing, semiconductor supply chains, and enterprise platform competition.

Anthropic – the company that makes the Claude AI assistant – disclosed on Monday, 01 Jun, that it has confidentially submitted a draft Form S-1 to the US Securities and Exchange Commission, taking the first formal step toward an initial public offering that could value the company above $1 trillion. The filing came days after Anthropic closed a $65 billion Series H financing round at a $965 billion post-money valuation, led by Altimeter Capital, Dragoneer, Greenoaks, and Sequoia Capital.

The valuation represents a more-than-doubling since February 2026, when Anthropic raised $30 billion at a $380 billion valuation. It also places Anthropic ~$113 billion above OpenAI's most recent reported valuation of $852 billion from its March 2026 round. Revenue has scaled at an extraordinary pace: the company generated $4.8 billion in Q1 2026, and projects $10.9 billion for Q2 – more than its entire 2025 annual revenue in a single quarter. The annualized run rate reached $47 billion by late May, driven primarily by Claude Code's rapid enterprise adoption. Anthropic's projected $559 million operating profit in Q2 would mark its first profitable quarter, though the ~5% margin remains thin for a company seeking a near-trillion-dollar public valuation. OpenAI is preparing a competing IPO targeting Q4 2026 at an estimated valuation near $1 trillion, setting up a direct race for public-market investor capital.

Microsoft – the technology company behind Windows, Azure, and the Office productivity suite – opened its Build 2026 developer conference today at Fort Mason Center in San Francisco. Microsoft AI chief Mustafa Suleyman unveiled MAI-Thinking-1, the company's first dedicated reasoning model built without model distillation – meaning it was not trained on outputs from another AI model. The model is targeted primarily at enterprise customers.

Beyond MAI-Thinking-1, Microsoft confirmed a broader slate of proprietary models. MAI-Image-2.5 debuted third on the Arena.ai text-to-image leaderboard with a score of 1,254 – a 72-point improvement over MAI-Image-2 and the first time a non-OpenAI, non-Google lab cracked the top five. MAI-Transcribe-1.5 builds on the April-released speech-to-text model that already claimed the lowest word error rate across 25 languages. MAI-Voice-2 adds support for 14+ languages including German, Spanish, French, Hindi, Japanese, Korean, and Portuguese. The conference is also expected to introduce a unified Copilot "super app" and updates to Azure AI Foundry – the company's enterprise AI platform – that formally add Anthropic Claude alongside OpenAI models with enterprise SLAs. The venue – capped at 2,500 attendees, the most intimate Build in years – and the timing alongside Computex underscore Microsoft's deliberate strategic pivot: building its own model stack to supplement, and potentially reduce reliance on, OpenAI.

Bloomberg reported on 01 Jun that Salesforce – the enterprise software company behind the Salesforce CRM platform – holds an ~$5 billion stake in Anthropic after participating in every funding round since Anthropic's Series C, starting with a $50 million initial investment in early 2023. Separately, Zoom – the video communications company – disclosed in its May regulatory filing that its Anthropic preferred shares are now worth ~$1.27 billion, based on the February 2026 valuation round – a return of ~$1 billion on an initial investment of $51 million made through Zoom Ventures in 2023, plus an additional $46 million added in Q1 2026.

Salesforce shares rallied ~10% on Monday, while Zoom surged ~11% to $112.86, its strongest single-day move in months. For Zoom, the $1.27 billion Anthropic stake now represents a significant portion of the company's ~$33.8 billion market capitalization and sits alongside $7.7 billion in cash with zero debt. Salesforce's broader strategic investment portfolio spans more than 450 companies with a combined carrying value of $7.8 billion as of 30 Apr – meaning the Anthropic position alone accounts for ~64% of that portfolio's total disclosed value. The 01 Jun Anthropic IPO filing would, if completed, give public markets the first opportunity to independently price these stakes – potentially moving them from illiquid footnotes to material balance-sheet drivers for both companies.

GitHub's – the Microsoft-owned code hosting and developer tools platform – transition from flat-rate subscription pricing to usage-based billing via "GitHub AI Credits" took effect on 01 Jun 2026. Under the new system, base plan prices remain unchanged – Free, Pro at $10/month, Pro+ at $39/month, Business at $19/user/month, Enterprise at $39/user/month – but chat, agent mode, and code review features now consume credits based on actual token usage (input, output, and cached tokens) at published API rates per model.

Code completions and Next Edit suggestions remain included in all plans without consuming credits. However, developers using agentic coding workflows – multi-step autonomous coding sessions where the AI writes, tests, and iterates on code independently – are reporting costs 10x to 50x higher than their previous monthly bills. GitHub has also retired annual plans; existing annual subscribers will remain on premium-request-based pricing until expiration, after which they move to Copilot Free unless they sign up for a new monthly paid subscription. The timing is strategic: the billing switch coincides with Build 2026, where Microsoft is announcing homegrown MAI coding models that could address the backlash by offering lower-cost alternatives within the Copilot stack.

Camtek Ltd. (NASDAQ: CAMT) – an Israeli company that makes automated optical inspection and metrology systems for the semiconductor industry – announced on 02 Jun that it has received multi-system orders totaling more than $105 million for AI-related semiconductor inspection and metrology equipment. The orders comprise a $55 million multi-system order from a tier-1 outsourced semiconductor assembly and test (OSAT) provider – a company that packages and tests chips on behalf of the designers – supporting AI applications, and over $50 million in Hawk system orders from a leading high-bandwidth memory (HBM) manufacturer – the specialized, high-speed memory components that power AI data centers – for AI-related applications. All systems are scheduled for delivery in 2027.

The orders build on an exceptionally strong first half for Camtek. Total Q1 2026 orders from leading OSATs alone exceeded $90 million, and a separate $45 million Hawk order from a tier-1 IDM (integrated device manufacturer – a company that designs and fabricates its own chips) was announced in February. Q1 2026 revenues reached $121.7 million (up 2.5% Y-o-Y from $118.6 million), and management guided Q2 revenue between $129 million and $131 million, with second-half 2026 revenues expected to grow by over 25% versus the first half. CEO Rafi Amit highlighted two structural themes: the strengthening of the OSAT business around 2.5D and 3D AI-related devices – advanced chip packaging techniques that stack multiple components for higher performance – and the Hawk platform's ability to meet the demanding inspection requirements of HBM manufacturing.

NVIDIA – the semiconductor company whose GPUs power most of the world's AI infrastructure – unveiled the RTX Spark Superchip at its Computex 2026 keynote in Taipei on 01 Jun. The chip pairs 20 Arm CPU cores with a Blackwell-class GPU featuring 6,144 CUDA cores – the parallel processing units that handle both graphics and AI calculations – connected via NVLink chip-to-chip interface, with up to 128 GB of unified LPDDR5X memory and 300 GB/s of memory bandwidth. NVIDIA is claiming 1 petaflop (one quadrillion operations per second) of AI compute, enabling local execution of models with up to 120 billion parameters.

Microsoft immediately announced the Surface Laptop Ultra as one of the first RTX Spark devices – a 15-inch, sub-18 mm, sub-2 kg machine positioned as a direct competitor to Apple's MacBook Pro. The device features a mini-LED PixelSense Ultra display, configurable up to 128 GB unified memory, full CUDA support, and native compatibility with Photoshop, Blender, DaVinci Resolve, and other professional creative applications. Pricing has not been disclosed; the device ships later in 2026. This marks the first time NVIDIA silicon has powered a Windows PC as its primary processor, ending Intel and AMD's decades-long duopoly in the consumer and professional laptop market. Anti-cheat software compatibility has been confirmed for League of Legends, Valorant, PUBG, and Alan Wake 2 – addressing one of the critical adoption barriers for Windows on Arm.

The AI coding tools market has reached $9.3 billion in 2026 and is projected to grow at 26% annually to ~$30 billion by 2031, according to Mordor Intelligence data cited in a CNBC analysis published 01 Jun. Anthropic's Claude Code – a command-line tool that lets developers delegate coding tasks to the Claude AI assistant from their terminal – has emerged as the market leader, with an annualized run rate exceeding $2.5 billion by early 2026 and business subscriptions quadrupling since the start of the year. Snowflake CEO Sridhar Ramaswamy told CNBC that his company's programmers primarily use the company's own CoCo tool alongside Claude Code.

The competitive landscape is intensifying on all fronts. Microsoft is unveiling proprietary MAI coding models at Build 2026 to power GitHub Copilot, reducing its reliance on OpenAI. Google – the technology company behind the Gemini AI model family – emphasized coding tools at its May I/O conference. OpenAI – the company behind ChatGPT – shifted its focus from consumer to enterprise, where its Codex offering directly targets Claude Code. Cursor – the AI-powered code editor – was valued at over $50 billion after raising $2 billion in April. Theory Ventures founder Tomasz Tunguz estimated that AI could eventually represent 30% to 60% of research and development spending. D.A. Davidson analyst Gil Luria described the coding market as "absolutely critical" for the major platforms, noting that the competition is not just about revenue but about locking developers into cloud ecosystems where training data flows back into model improvement.