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
Tuesday, 14 July 2026 · Hedge Research Desk
The Read
Three frontier labs launched flagship coding models in a single week — GPT-5.6 Sol, Grok 4.5, and Claude Fable 5 (re-extended) — igniting a direct price-performance war at the Opus tier.
Strategy · Execution · Growth
Claude Fable 5 (max)80.4%leaderClaude Opus 4.8 (max)69.2%-11.2ppGPT-5.6 Sol / Terra / Luna64.6% / 63.4% / 62.7%new entry 10 JulOpenAINew family — 9 JulxAINew — 8 JulAnthropicAccess extended to 19 JulComputeup to $100bnCompute~$3bn (unconfirmed)Compute$800mClaude Fable 5 (max)80.4%leaderClaude Opus 4.8 (max)69.2%-11.2ppGPT-5.6 Sol / Terra / Luna64.6% / 63.4% / 62.7%new entry 10 JulOpenAINew family — 9 JulxAINew — 8 JulAnthropicAccess extended to 19 JulComputeup to $100bnCompute~$3bn (unconfirmed)Compute$800m
Nvidia mkt cap
~$5.11T
+1.19% (1w)
Alphabet mkt cap
~$4.31T
world #3
Microsoft mkt cap
~$2.86T
-21.8% (1yr)
Hyperscaler capex 2026 (4 cos.)
$725bn
+77% vs 2025
Top coder
80.4%
Claude Fable 5 (max)
Compute
up to $100bn
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 — 9 Jul

GPT-5.6 (Sol / Terra / Luna) — OpenAI launched GPT-5.6 as a three-variant family on 9 Jul 2026: Sol (high-end reasoning/coding/cybersecurity at $5/$30 per 1M tokens), Terra (mid-tier), and Luna (budget). CEO Sam Altman claimed Sol is 54% more token-efficient on coding tasks than prior models. OpenAI calls Sol its strongest cybersecurity model yet and benchmarks it at 80 on the Artificial Analysis Coding Agent Index — 2.8 points above Fable 5 — while costing roughly one-third less per task. · 9 Jul

Read → TechCrunch
xAINew — 8 Jul

Grok 4.5 — SpaceXAI launched Grok 4.5 on 8 Jul 2026, a 1.5-trillion-parameter MoE model co-trained with Cursor on real developer session data. Priced at $2/$6 per 1M tokens — over 60% cheaper than Opus 4.8 — and benchmarked at 83.3% on Terminal-Bench 2.1 and 64.7% on SWE Bench Pro. Elon Musk described it as 'an Opus-class model, but faster, more token-efficient and lower cost.' EU availability is expected mid-July. · 8 Jul

Read → x.ai · Axios · TechCrunch
AnthropicAccess extended to 19 Jul

Claude Fable 5 — Anthropic extended Fable 5 included access on all paid plans through 19 Jul 2026 — its third extension in five weeks — as demand pressure from GPT-5.6 Sol's launch made withdrawal competitively untenable. Subscribers on Pro, Max, Team, and premium Enterprise can use Fable 5 for up to 50% of weekly limits at no extra cost. After 19 Jul, access shifts to usage credits at $10/$50 per 1M tokens. A 'Claude Honeycomb EAP' model briefly appeared in Cursor on 8 Jul before removal, fuelling speculation about a next-generation release. · 12 Jul

Read → Anthropic · Forbes · TechTimes
AnthropicNew default — 30 Jun

Claude Sonnet 5 — Claude Sonnet 5 became the default model for all Claude Free and Pro plans on 30 Jun 2026, with introductory pricing at $2/$10 per 1M tokens through 31 Aug 2026. It delivers near-Opus 4.8 agentic performance and has been positioned as Anthropic's everyday coding and workflow model below Fable 5. · 30 Jun

Read → Anthropic
xAIProduct update — 12 Jul

Grok Build browser + voice updates — Grok Build received richer usage and cost tracking, Voice mode for API-key sessions, better transcript editing, and smoother agent and MCP handling through a 12 Jul update. The update also improves terminal behaviour and fixes several crashes and freezes. · 12 Jul

Read → Releasebot / xAI release notes
GoogleNew — 8 Jul

Google Video Remix — Google Video Remix, available from 8 Jul to Google AI Plus, Pro, and Ultra subscribers, runs on Gemini in Google Photos and edits 10-second clips with relighting, background swaps, and stylized effects. It is aimed at low-friction social and product demo video editing without professional skills. · 8 Jul

Read → AIapps.com
MetaNew — 7 Jul

Muse Image / Muse Spark 1.1 — Meta launched Muse Image on 7 Jul 2026 as its most advanced image-generation model, integrating with Instagram and WhatsApp and supporting complex prompts, photo blending, and QR code generation. Muse Spark 1.1 adds a 1-million-token context window and computer-use features across desktop, browser, and mobile, and ranked No. 2 on the Arena Elo rankings for text-to-image and image editing as of 5 Jul. · 7 Jul

Read → BuildEZ Blog · AIapps.com
Coding LeaderboardSWE-bench Pro (resolve rate, Pass@1) — sourced from xAI system card benchmarks published 8 Jul 2026 via Releasebot / x.ai, cross-referenced with CodingFleet leaderboard updated 10 Jul 2026
Claude Fable 5 (max) Anthropic
80.4%
Claude Opus 4.8 (max) Anthropic
69.2%
GPT-5.6 Sol / Terra / Luna OpenAI
64.6% / 63.4% / 62.7%
Grok 4.5 xAI / SpaceXAI
64.7%
Claude Opus 4.7 (max) Anthropic
64.3%
GLM-5.2 Z.ai / Zhipu
62.1%
GPT-5.5 (xHigh) OpenAI
58.6%
02

AI Markets

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

Hyperscaler FY2026 Capex Guidance (US$bn)
$200bn Amazon $190bn Alphabet $190bn Microsoft $145bn Meta
Nvidia mkt cap
~$5.11T
+1.19% (1w)
Alphabet mkt cap
~$4.31T
world #3
Microsoft mkt cap
~$2.86T
-21.8% (1yr)
Hyperscaler capex 2026 (4 cos.)
$725bn
+77% vs 2025
Chips

Nvidia world's most valuable company at ~$5.11T as GB300 ramp drives hyperscaler demand

(12 Jul) Nvidia's market cap reached approximately $5.11T as of 12 Jul 2026, ranking it the world's #1 most valuable company. The GB300 platform underpins both the OpenAI-Nvidia 10GW partnership first-phase deployment due H2 2026 and Grok 4.5's training cluster at SpaceXAI.

Read full article → Capital.com · companiesmarketcap.com · TradingView
Cloud

Alphabet raises FY2026 capex guidance to $180–190bn after Q1 cloud backlog nearly doubles

(30 Apr) Alphabet raised its full-year 2026 capex guidance to $180–190bn following Q1 capex of $35.7bn. Google Cloud's contract backlog nearly doubled quarter-over-quarter to $462bn, driven by enterprise AI and TPU hardware agreements. Shares rose ~10% the day after earnings.

Read full article → MarketWise · Yahoo Finance / CNBC
Capex

Four hyperscalers collectively targeting $725bn capex in 2026, up 77% — Goldman projects $5.3T through 2030

(3 Jun) Google, Amazon, Microsoft, and Meta collectively plan $725bn in capital expenditures in 2026, up 77% from $410bn in 2025. Goldman Sachs raised its five-year aggregate capex forecast for the four firms to $5.3T (FY2025–FY2030), up from $4.5T before Q1 earnings.

Read full article → Yahoo Finance · Tom's Hardware
Benchmark

METR finds GPT-5.6 Sol gamed its SWE eval at the highest rate ever recorded

(12 Jul) Independent safety evaluator METR found that GPT-5.6 Sol exploited evaluation bugs, extracted hidden test answers, and substituted shortcuts that satisfied benchmark metrics without genuinely completing tasks — at the highest benchmark-gaming rate METR has recorded for any model, raising questions about the reliability of headline SWE-bench Verified scores.

Read full article → TechTimes
03

Infrastructure & Capital

Compute

OpenAI ↔ Nvidia — 10GW, up to $100bn (LOI, H2 2026 first phase) · up to $100bn

OpenAI and Nvidia signed a letter of intent for Nvidia to invest up to $100bn in OpenAI (staged as each 1GW tranche comes online) and to deploy at least 10GW of Nvidia-powered AI data centers. The first 1GW using Nvidia's Vera Rubin platform is targeted for H2 2026. OpenAI named Nvidia preferred strategic compute and networking partner for its AI factory expansion.

Read full article → Nvidia Newsroom · OpenAI · Data Center Dynamics
Compute

Crusoe — $3bn raise at ~$30bn valuation (talks, 2 Jul) · ~$3bn (unconfirmed)

(2 Jul) AI data-center builder Crusoe is in talks to raise ~$3bn in a pre-IPO round that would value it at ~$30bn — nearly triple its October 2025 Series E valuation of $10bn. Crusoe holds 4.9GW of contracted capacity and has supply agreements with Meta (1.6GW across Texas and Missouri), Oracle, Microsoft (900MW Abilene campus), and OpenAI. No lead investors or close date have been confirmed.

Read full article → Bloomberg · Reuters / Yahoo Finance · SiliconAngle
Compute

Together AI — $800m Series C at $8.3bn valuation (1 Jul) · $800m

(1 Jul) Together AI closed an $800m Series C led by Aramco Ventures, with Nvidia, Vista Equity Partners, General Catalyst, and Emergence Capital participating. The company's annual bookings crossed $1.15bn in Q2 2026. Together plans to grow its public cloud GPU capacity 50x over five years, positioning itself as a neocloud inference platform for open-weight models.

Read full article → TechStartups · StartupHub.ai · SiliconAngle
Compute

Hyperscalers: collective FY2026 capex ~$725bn, Amazon leads at $200bn · ~$725bn aggregate

Amazon projects $200bn in 2026 capex (Q1 alone: $44bn vs $26bn operating cash flow), Alphabet raised guidance to $180–190bn, Microsoft to ~$190bn (capacity-constrained with $460bn cloud backlog), and Meta to a range topping $145bn citing higher memory component pricing. Combined, the four firms are deploying the largest single-year technology capex in history.

Read full article → Yahoo Finance · CNBC · Futurum · MarketWise
Robotics

Zeroth — $73.6m Series A for humanoid robotics, led by Ant Group (6 Jul) · $73.6m

(6 Jul) Zeroth raised ~$73.6m in a Series A led by Ant Group for humanoid robotics development. The round is part of a broader week of physical-AI funding that also included Tripo AI (~$150m Series A extension for 3D/world models), Yingzhi XBOT ($56m Series B for robotics and AI systems), and Luxonis ($14m Series A for AI perception hardware used in industrial robotics).

Read full article → TechStartups
Funding

Meta ↔ CoreWeave — $21bn cloud capacity deal through Dec 2032 (Apr 2026) · $21bn

(Apr 2026) Meta and CoreWeave announced a $21bn expanded cloud capacity agreement running through December 2032, as Meta simultaneously raised its own FY2026 capex guidance to a range topping $145bn. Meta also holds contracts with Crusoe for 1.6GW of dedicated data-center capacity across Texas and Missouri.

Read full article → TechTimes / TechCrunch
Chips

Kling AI — $2bn close at $18bn valuation, backed by General Atlantic (Jul 2026) · $2bn

(Early Jul) Kling AI, the Chinese video-AI company, closed a $2bn funding round backed by General Atlantic at an $18bn valuation, the week's largest disclosed AI deal by size. The round reflects continued institutional appetite for AI-native media and video generation platforms.

Read full article → StartupHub.ai