Tuesday AI Insights 🔴 Live Today AI Strategy UAE Agentic AI Geopolitics

The Week AI Got Honest — And Then the Hardware Wars Started

Anthropic hit $900B. Sam Altman admitted he was wrong. DeepSeek permanently reset what AI should cost. The UAE broke 70.1%. Yesterday Nvidia announced it is now a PC company and Microsoft cut its OpenAI dependency. Today Computex Day 2 and Microsoft Build Day 2 are live. Here is everything verified, dated, and analysed — without the hype.

LW
Lisa Warren
Founder & CEO, Neural Horizons AI  ·  2 June 2026  ·  12 min read
Editorial note on sourcing: This edition covers three time windows, clearly labelled throughout. Sections 01–06 are last week's verified news (25–31 May 2026). Sections 07–08 cover yesterday's major launches (1 June 2026) from Computex and Microsoft Build Day 1. Section 09 covers what is live today, 2 June 2026. Background context from earlier months is explicitly labelled as such — never as current news.

⬡ 00 — Key Takeaways · Read this first

Eight things that changed in the last eight days

👑

Anthropic is now the world's most valuable private AI company. $30B+ closed at above $900B pre-money the week of 26 May — surpassing OpenAI's March valuation. The pace from $61.5B in March 2025: 14 months.

🎙️

Sam Altman admitted he was "pretty wrong" about AI jobs — four days after OpenAI filed for IPO. Dario Amodei said the same week. Fortune called it a coordinated walk-back timed to capital markets. The truth sits in the middle.

💥

DeepSeek's 75% V4-Pro price cut went permanent on 22 May. Frontier-grade AI is now 11× cheaper from China than from OpenAI. This is the new floor — and it changes every AI budget conversation.

🇦🇪

The UAE hit 70.1% AI adoption — the first economy on earth to break 70%. Per the Microsoft AI Diffusion Report Q1 2026. Global average: 17.8%. But the statistic has two sides, and only one of them is good news.

🌍

Anthropic and the Gates Foundation committed $200M to deploy Claude in global health, education, and economic mobility — targeting 4.6 billion people over four years.

🤖

Gemini 3.5 Flash went GA last week — beating Gemini 3.1 Pro on coding and agentic benchmarks while running 4× faster. Google's message: Flash is no longer the budget option. Flash is the flagship.

🔴

Yesterday: Nvidia launched the RTX Spark Superchip at Computex. Arm-based, co-developed with Microsoft — Blackwell GPU, 20-core Grace CPU, 128GB unified memory, 1 petaflop on-device AI. Dell, HP, Asus, Lenovo, MSI, Surface arriving this autumn. Intel fell 6%. Nvidia is now a PC company.

🔴

Yesterday: Microsoft Build revealed Project Polaris — Microsoft's own AI coding model replacing GPT-4 in GitHub Copilot by August 2026. Windows Agent Framework open-sourced. Microsoft is designing for a post-OpenAI-dependency future.

Last week was one of those rare moments when the AI industry paused the performance of confidence and offered something closer to honesty — selectively, carefully, with one eye always on the capital markets. That makes it more interesting, not less. The job of Tuesday AI Insights is not to amplify the narrative. It is to ask what is actually true, what is being strategically framed, and what it means for leaders building in this region.

⬡ Last Week · Verified 25–31 May 2026

01 — Anthropic Overtakes OpenAI: $900B, $30B Raised, and an Enterprise Cost Crisis Buried in the Footnotes

✓ Confirmed: week of 26 May 2026

The round closed during the week of 26 May. Sequoia Capital, Dragoneer, Altimeter, and Greenoaks co-led, each contributing approximately $2 billion. Peter Thiel's Founders Fund, General Catalyst, Microsoft, and Nvidia also participated. The total exceeded $30 billion at a pre-money valuation above $900 billion — making Anthropic the world's most valuable private AI company, ahead of OpenAI's $852 billion March valuation. [Bloomberg / Yahoo Finance]

$900B+
Pre-money valuation at close
$30B
New capital — largest round in Anthropic history
$30B
Annualised revenue run rate, April 2026
14 mo
From $61.5B to $900B+ — pace of ascent

The revenue story is real. Enterprise customers now represent around 80% of revenue, with more than 1,000 businesses spending over $1 million annually. Claude Code drove significant growth — which makes the detail that Microsoft quietly cancelled most of its Claude Code licences last week especially worth noting.

⚠ The signal buried in the headline The Verge reported last week that Microsoft cancelled most of its Claude Code licences — redirecting engineers to GitHub Copilot CLI — after per-engineer token bills ran between $500 and $2,000 per month. Uber's COO confirmed the company burned through its entire 2026 AI budget by April with no clear line to improved outcomes. Anthropic's revenue growth is real. So is the cost reckoning happening inside its biggest enterprise customers. [Inc]

"No company in American technology history has grown at this rate. The question is whether the infrastructure built to justify these valuations holds under enterprise cost scrutiny."

— The Next Web, May 2026

02 — Sam Altman Said He Was "Pretty Wrong." Dario Amodei Agreed. Both Have IPOs Coming.

✓ Confirmed: 26 May 2026

On 26 May, speaking in Sydney, Sam Altman told the audience he was "pretty wrong" about AI's economic impact on jobs. He had expected more entry-level white-collar displacement by now. Dario Amodei — who once predicted AI could eliminate 50% of white-collar jobs — said the same week that automation may actually expand the work people do. [Fortune, 26 May]

OpenAI filed its S-1 confidentially on 22 May. Altman's reversal came four days later. Anthropic is eyeing its own listing. Calling this a coincidence requires more faith than the evidence supports. [SpaceDaily]

⬡ Seeing both sides

What is true
Yale Budget Lab data through March 2026 shows no meaningful change in unemployment for high AI-exposure occupations. AI is replacing tasks within jobs — not yet entire roles at scale.
What is also true
Meta laid off 8,000 employees last week and cited AI. 99% of C-level execs in a Mercer survey expect AI to reduce headcount within two years. Agentic AI products are weeks old — the displacement data does not exist yet. [Yahoo Finance]

"The organisations using this window to embed AI into operations will be the ones setting the baseline when the curve steepens."

— Lisa Warren, Neural Horizons AI

03 — DeepSeek Made the 75% Price Cut Permanent. This Is Not a Promotion. This Is the New Floor.

✓ Confirmed: 22 May 2026

On 22 May, DeepSeek announced its promotional discount on V4-Pro — due to expire 31 May — would not roll back. It is now the permanent list price: $0.435 per million input tokens, $0.87 per million output tokens. OpenAI's GPT-5 charges $2.50 input and $10 output. DeepSeek V4-Pro is now 11× cheaper on input, 34× cheaper on output. [The Next Web, 24 May]

–75%
V4-Pro price cut — permanent from 22 May
11×
Cheaper than GPT-5 on input tokens
60%
OpenRouter API traffic already from Chinese models
$0.87
New permanent output price per million tokens
⬡ What this means for your AI budget Any AI infrastructure budget built on 2025 pricing is now out of date. If your team routes significant workloads through frontier models, the question of whether lower-cost alternatives can handle those tasks without meaningful quality loss is now a leadership conversation — not just a technical one. This is not a recommendation to switch. It is a prompt to know what you are paying for and why. [CTOL Digital]

04 — The UAE Hit 70.1% AI Adoption. Here Is What That Confirms — and What It Still Doesn't Tell You.

✓ Microsoft AI Diffusion Report Q1 2026, published mid-May 2026

The Microsoft AI Economy Institute placed the UAE first globally: 70.1% of working-age people now use AI tools — the first economy anywhere to cross 70%. The global average is 17.8%. [Gulf News]

⬡ Both sides of 70.1%

What it confirms
The UAE's coordinated approach — UAE AI Strategy, National AI Officer network, mandatory government upskilling — has produced real, measurable penetration. The government-private sector alignment is working as designed.
What it does not confirm
Adoption measures tool use — not changed outcomes. The UAE's 54% adoption rate earlier this year coexisted with most businesses failing to convert usage into measurable outcomes. A higher number makes closing the execution gap more urgent — not less. Read our full analysis →

The UAE also launched the Cyber Factory — a sovereign AI-driven cybersecurity initiative — in mid-May, as the Cyber Security Council reported over 800,000 attacks per day. As AI adoption rises, so does the AI-powered attack surface. [Computer Weekly]


05 — Anthropic and the Gates Foundation: $200M to Deploy Claude Where the Market Won't Go

✓ Confirmed: 14 May 2026

Anthropic and the Bill & Melinda Gates Foundation announced a four-year, $200 million partnership to deploy Claude across global health, life sciences, education, and economic mobility. Target: 4.6 billion people in low- and middle-income countries lacking access to essential health services. [Anthropic, 14 May]

◦ What the partnership covers Global health: screening drug and vaccine candidates for polio, HPV, eclampsia; improving deployment forecasts for malaria and tuberculosis. Education: K-12 tutoring, college advising, literacy and numeracy with public benchmarks. Agriculture: crop datasets for smallholder farmers. Economic mobility: portable skills records.

06 — Gemini 3.5 Flash: Google Stops Treating Flash as the Budget Option

✓ Launched 19 May at Google I/O · GA confirmed 29 May 2026

Google launched Gemini 3.5 Flash at I/O on 19 May and confirmed GA last week. It beats Gemini 3.1 Pro on coding and agentic benchmarks while running 4× faster. API pricing: $1.50 input / $9 output per million tokens. It is now the default model in the Gemini app, Google Search's AI Mode globally, and the Gemini API. [TechCrunch, 19 May]

⬡ The strategic signal Flash is now the default tier for agentic-first developers — not the budget fallback. Google also launched Managed Agents in the Gemini API: stateful agents running in secure, Google-hosted Linux sandboxes. The agentic pivot is not coming. It has arrived. Gemini 3.5 Pro — 2M token context, Deep Think reasoning — is delayed to this month. Sundar Pichai told I/O: "give us until next month." [Google AI Dev Docs]
⬡ Yesterday · 1 June 2026 · Computex & Microsoft Build Day 1

07 — Nvidia Just Became a PC Company: The RTX Spark Superchip Changes On-Device AI

⬡ Yesterday · 1 June 2026 · Computex 2026, Taipei

In Taipei yesterday, Jensen Huang unveiled the RTX Spark Superchip — an Arm-based PC processor co-developed with Microsoft and MediaTek. Blackwell RTX GPU with 6,144 CUDA cores, 20-core Grace CPU, up to 128GB unified memory, one petaflop of AI performance, built on TSMC's 3nm process. Devices from Dell, HP, Asus, Lenovo, MSI, and Microsoft Surface arriving this autumn. Intel fell 6% on the announcement. [CNBC]

6,144
CUDA cores — Blackwell RTX GPU
128GB
Unified memory — top config
1 PFLOP
On-device AI performance
–6%
Intel stock on announcement
⬡ What this means for MENA businesses One petaflop of AI performance on-device means a meaningful class of agentic workloads no longer needs to leave the device — or the country. For organisations thinking about data sovereignty and on-premise AI, the RTX Spark changes the conversation. Local inference at this scale in mainstream laptops this autumn is architectural, not incremental. [Tom's Hardware]

08 — Microsoft Build Day 1: Project Polaris — Microsoft Is Building Its Own AI to Replace GPT-4 in Copilot

⬡ Yesterday · 1 June 2026 · Microsoft Build, San Francisco

Microsoft unveiled Project Polaris — its own in-house AI coding model replacing GPT-4 Turbo as the default for GitHub Copilot from August 2026, with automatic migration and a three-month optional fallback. Microsoft also open-sourced the Windows Agent Framework, announced the Azure Agent Mesh, and shipped Copilot Workspace out of beta. [ChatForest recap]

⬡ Both sides of Microsoft's OpenAI independence move

What it means for Microsoft
Project Polaris gives Microsoft cost control, IP ownership, and strategic independence from OpenAI — whose IPO will fundamentally change the commercial relationship. After cancelling most Claude Code licences last week and now building its own model, Microsoft is designing for a post-OpenAI-dependency future.
What it means for everyone else
GitHub Copilot had a 9% developer love score in the JetBrains April survey — against 46% for Claude Code. If Polaris doesn't close that gap quickly, Microsoft's developer credibility takes a lasting hit. The Windows Agent Framework open-sourcing directly challenges Google's Antigravity and Anthropic's agent APIs. [Notebookcheck]
🔴 Live Today · 2 June 2026 · Computex Day 2 & Microsoft Build Day 2

09 — Live Today: Intel Enters the AI GPU Race, Surface Laptop Ultra Revealed, Jensen Huang at Build

🔴 Live: 2 June 2026 · Computex Day 2 + Microsoft Build Day 2

Computex 2026 officially opened today, running through 5 June in Taipei under the theme "AI Together." Microsoft Build Day 2 runs simultaneously in San Francisco. [Tom's Hardware live]

🔴 Breaking from Computex today — Intel Crescent Island AI GPU Intel has unveiled its Crescent Island data-centre GPU — built on Xe3P architecture, featuring up to 480GB of LPDDR5X memory, explicitly positioned as "built for agentic AI." This is Intel's most significant AI GPU announcement in years and its direct answer to Nvidia's data-centre dominance. [Tom's Hardware, 2 June]
🔴 Breaking from Computex today — Microsoft Surface Laptop Ultra Microsoft has unveiled the Surface Laptop Ultra — the most direct MacBook Pro competitor the company has ever shipped. It carries Nvidia's RTX Spark Superchip: 20-core CPU, Blackwell GPU, 128GB unified RAM, in a 15-inch chassis with mini-LED display, HDMI, USB-C, USB-A, and SD card reader. [Tom's Hardware, 2 June]
🔴 Live — Microsoft Build Day 2 Jensen Huang joins the Build stage today. Day 2 sessions focus on Azure AI Foundry enterprise integrations, multi-agent orchestration, and the developer SDK for Windows on-device AI. AMD's Radeon RX 9070 GRE also goes global today at $549.

10 — What to Watch This Week

⬡ On the radar — week of 2 June 2026

Today
Microsoft Build Day 2 + Computex Day 2 — Jensen Huang on stage at Build. Intel Crescent Island and Surface Laptop Ultra announced. Watch the Azure AI Foundry cost governance sessions and the Windows AI SDK. [Live blog →]
3 June
Neural Horizons AI Webinar — Live executive briefing with Lisa Warren on closing the UAE execution gap. Register free →
2–5 June
Computex 2026 continues through Friday. AMD Nova Lake CPU teasers expected. Watch the budget PC segment — Apple's MacBook Neo is forcing every OEM to compete on AI capability simultaneously.
June 2026
Gemini 3.5 Pro GA expected — Sundar Pichai said "give us until next month." 2M-token context, Deep Think reasoning, targeting Claude Opus 4.7 and GPT-5.5 head-to-head.
8–12 June
Apple WWDC 2026 — New CEO John Ternus must show Siri 2.0 working as a genuine on-screen AI agent. Apple's on-device strategy now has direct hardware competition from Nvidia's RTX Spark and Intel's Crescent Island.
30 June
Colorado AI Act takes effect — first US state-level comprehensive AI regulation on high-risk systems. If your business serves US customers with AI-powered decision-making, compliance planning is non-optional.
24 July
DeepSeek alias retirement — migrate production code from legacy deepseek-chat or deepseek-reasoner to deepseek-v4-pro before 24 July or face API failures.

Insights Are Not Strategy

Tuesday AI Insights tells you what happened. A Neural Horizons AI engagement tells you what to do about it — built for leaders operating in this region.

Transparent pricing. Proven frameworks. Measurable outcomes.

Lisa Warren is the Founder & CEO of Neural Horizons AI, a boutique AI consultancy based in Dubai's Meydan Free Zone. She advises SMEs and mid-market companies across the UAE and MENA region on AI strategy, leadership, and implementation. Connect on LinkedIn.

Related Insights

Sources