Services01Investment & Deal Advisory02Project Technical Advisory03Governance, Risk & Compliance04Fintech & Payments Licensing05Startup & Venture Support06Technology & AI Transformation
Case Studies Insights Hedge Learn Hedge Labs About Us Contact Us

AI & Technology Update

Frontier models · Markets · Infrastructure · Capital
Sunday, 12 July 2026 · Hedge Research Desk
The Read
A 72-hour release sprint — OpenAI GPT-5.6 (9 Jul) and xAI Grok 4.5 (8 Jul) go head-to-head on coding benchmarks while hyperscaler capex commitments cross $725 bn for 2026.
Strategy · Execution · Growth
Claude Mythos 595.5%—Claude Fable 595.0%—Claude Opus 4.888.6%—OpenAINew familyxAINew flagshipAnthropicNew defaultCompute$800MCompute$40bn combinedCompute$1.2bnClaude Mythos 595.5%—Claude Fable 595.0%—Claude Opus 4.888.6%—OpenAINew familyxAINew flagshipAnthropicNew defaultCompute$800MCompute$40bn combinedCompute$1.2bn
Nvidia mkt cap
~$4.91tn
+3.5% 1-wk
Alphabet mkt cap
~$4.34tn
—
Microsoft mkt cap
~$2.83tn
—
Mag-7 S&P 500 weight
32.5%
steady
Top coder
95.5%
Claude Mythos 5
Compute
$800M
deal
01

Frontier Models

The week across the frontier labs — new defaults, previews and open-weight releases — plus where the coding benchmarks stand.

OpenAINew family

GPT-5.6 (Sol / Terra / Luna) — OpenAI launched GPT-5.6 on 9 Jul in three tiers: Sol (workhorse, $5/$30 per M tokens), Terra (intermediate), and Luna (budget, $1/$6). CEO Altman claimed Sol is 54% more token-efficient on AI coding tasks and called it 'the strongest cybersecurity model yet.' The family directly targets Anthropic's Fable 5 on the Artificial Analysis Coding Agent Index, where OpenAI claims Sol scores 80 — 2.8 points above Fable 5. · 9 Jul

Read → TechCrunch
xAINew flagship

Grok 4.5 — SpaceXAI released Grok 4.5 on 8 Jul — its first model jointly trained with Cursor session data and built on a 1.5T V9 MoE foundation across tens of thousands of NVIDIA GB300 GPUs. Priced at $2/$6 per M tokens (>60% cheaper than Claude Opus 4.8), it achieves 83.3% on Terminal-Bench 2.1 and resolves SWE-bench Pro tasks using ~4.2× fewer output tokens than Opus 4.8. Available in Grok Build, Cursor (all plans), and the SpaceXAI API; EU availability expected mid-July. · 8 Jul

Read → SpaceXAI official · Axios · TechCrunch
AnthropicNew default

Claude Sonnet 5 — Anthropic launched Claude Sonnet 5 on 30 Jun as its most agentic Sonnet yet, replacing Sonnet 4.6 as the Free and Pro default. It brings near-Opus-4.8 performance (63.2% on agentic coding vs. Opus 4.8's 69.2%) at ~1/3 the flagship cost. Introductory pricing: $2/$10 per M tokens through 31 Aug 2026, then $3/$15. Features a 1M-token context window and is live in Claude Code, Cursor, VS Code, and GitHub Copilot. Anthropic simultaneously launched Claude Science, a desktop app for scientific research workflows. · 30 Jun

Read → Anthropic
KimiOpen weights / Modified MIT

Kimi K2.7 Code — Moonshot AI shipped Kimi K2.7 Code on 12 Jun, posting +21.8% improvement over K2.6 on Kimi Code Bench v2. A HighSpeed variant delivers approximately 6× faster multimodal coding inference. The model is available on GitHub Copilot as the first open-weight option on that platform, giving cost-aware teams a cheaper usage-based billing path. · 12 Jun

Read → LLM Gateway · Taskade / Fello AI
Coding LeaderboardSWE-bench Verified (BenchLM.ai, last updated 8 Jul 2026)
Claude Mythos 5 Anthropic
95.5%
Claude Fable 5 Anthropic
95.0%
Claude Opus 4.8 Anthropic
88.6%
GPT-5.5 OpenAI
88.7%
Claude Opus 4.7 Anthropic
87.6%
GPT-5.3 Codex OpenAI
85.0%
02

AI Markets

AI leaders and their capital cycle — capex commitments alongside the month’s market moves.

Hyperscaler capex — FY2026 guidance (US$bn)
$200bn Amazon $190bn Microsoft $180bn Alphabet $125bn Meta
Nvidia mkt cap
~$4.91tn
+3.5% 1-wk
Alphabet mkt cap
~$4.34tn
—
Microsoft mkt cap
~$2.83tn
—
Mag-7 S&P 500 weight
32.5%
steady
Chips

Nvidia expands heterogeneous AI chip partnership strategy

(9 Jul) Nvidia is increasingly embracing partnerships with emerging AI chip developers, signaling a shift from competing solely on hardware to enabling heterogeneous AI infrastructure, per The Information / Digitimes.

Read full article → Digitimes
Chips

Nvidia Kyber next-gen rack system faces reported delay to 2028

(10 Jul) Reports surfaced that Nvidia's next-generation Kyber AI rack system may slip to 2028 due to manufacturing snags; Nvidia denied the report and said its roadmap is intact, per CNBC and Yahoo Finance.

Read full article → Distill Intelligence chip briefing
Chips

Broadcom & OpenAI unveil 'Jalapeño' custom AI processor

(Jul) Broadcom and OpenAI unveiled 'Jalapeño', OpenAI's first custom intelligence processor, while Apple extended its multi-billion-dollar wireless chip supply agreement with Broadcom through 2031.

Read full article → Distill Intelligence chip briefing
Compute

Four hyperscalers to spend $725bn on capex in 2026 — up 77% YoY

(Feb–Jun 2026, running story) Amazon ($200bn), Microsoft (~$190bn), Alphabet ($175–185bn), and Meta ($115–135bn) collectively target $725bn in 2026 capex, nearly doubling 2025's $410bn, with ~75% earmarked for AI data centers and GPU clusters. Amazon faces projected negative free cash flow of ~$17bn as a result.

Read full article → Yahoo Finance / Goldman Sachs analysis · CNBC · Futurum Research
03

Infrastructure & Capital

Compute

Together AI — $800M Series C at $8.3bn valuation · $800M

(1 Jul) Aramco Ventures led an $800M Series C in Together AI (GPU-cloud neocloud), with NVIDIA, Vista Equity Partners, General Catalyst, and Emergence Capital participating. Annual bookings crossed $1.15bn in Q2 2026. Together plans to expand public cloud capacity 50× over five years.

Read full article → TechStartups · Crunchbase
Compute

Nvidia stakes — $30bn in OpenAI, $10bn in Anthropic · $40bn combined

(Early 2026, confirmed) Nvidia finalized a $30bn investment in OpenAI and a $10bn stake in Anthropic in early 2026, its two largest AI lab bets ever, before CEO Jensen Huang signaled these would likely be Nvidia's last mega-checks to frontier labs as lab rivalry turned public. Anthropic's May 2026 Series H valued it at $965bn.

Read full article → AI Funding Tracker
Compute

Quantum Systems — $1.2bn Series D at $8bn valuation · $1.2bn

(2 Jul) European defense AI and drone-autonomy firm Quantum Systems raised $1.2bn Series D co-led by Blackstone and Airbus, doubling its valuation to $8bn. The company cited 19,000+ missions flown in Ukraine and plans to expand its MOSAIC UXS software across air, land, and maritime systems.

Read full article → TechStartups · Bloomberg
Funding

Global VC hits record $510bn in H1 2026 · $510bn H1 2026

(Jul) Crunchbase reports that global venture capital reached a record $510bn in H1 2026 — a new high for any half-year — driven overwhelmingly by AI-related deals, with capital concentrating in AI infrastructure, robotics, and domain-specific AI at late-stage and growth rounds.

Read full article → TechStartups citing Crunchbase
Compute

Joulent — $1.75bn energy infrastructure financing · $1.75bn

(Jun 27–Jul 2) Houston-based Joulent, focused on AI compute-intensive energy infrastructure, secured the week's largest disclosed round at $1.75bn strategic financing, reflecting intensifying demand for AI-adjacent power capacity.

Read full article → Crunchbase News
Robotics

Zeroth — $73.6M Series A for humanoid robotics · $73.6M

(6 Jul) Zeroth raised approximately $73.6M in Series A funding for humanoid robotics, led by Ant Group, as physical AI deployment continues to attract serious capital alongside agentic systems.

Read full article → TechStartups
Funding

Tripo AI — ~$147–150M Series A extension for 3D/world models · ~$150M

(Early Jul) Tripo AI, a California-based generative AI startup for 3D and world models with robotics applications, raised approximately $147–150M in a Series A extension with investors including 4399.com.

Read full article → TechStartups