The AI infrastructure battle is now physical — Bezos, Blackstone, and Samsung

Advertise here | 6-min Read

50 Excel Hacks That Make Spreadsheets Work For You

Most Excel users waste hours on tasks that take experts minutes. The difference is not talent. It is knowing the right shortcuts.

These 50 Excel hacks from Kenji Explains cover the functions, formulas, and workflows that make spreadsheets work for you instead of against you.

Each one comes with step-by-step instructions and ready-to-use templates. Subscribe to Marketing Against the Grain and get all 50 free.

Here's what you'll get:

• Time-saving shortcuts that eliminate the most common formula frustrations

• Interface essentials most users never discover on their own

• Game-changing functions for data analysis, automation, and visualization

• Guided practice template so you can apply every hack immediately

The people who are fast in Excel are not smarter. They just stopped fighting the tool.

Get 50 Excel Hacks free when you subscribe to Marketing Against the Grain today.

Get The Guide

Top news of the day

1. Kimi K3 demand crashed its own infrastructure — new subscriptions paused, existing users unaffected: Moonshot AI has temporarily stopped accepting new subscriptions for its Kimi K3 model after demand pushed the company's available computing capacity to near capacity. The pause highlights a structural problem facing Chinese AI developers: access to the most advanced Nvidia chips remains constrained — the same export controls that were supposed to slow Chinese AI development are now also preventing Chinese labs from scaling the models that Western infrastructure teams cannot match on price. Kimi K3 is currently available in the app and API for existing users. Full weights still drop July 27. The demand signal is unambiguous: the market for frontier-adjacent open-weight models at competitive pricing is larger than anyone planned for.

2. Chip stocks just had their worst week since April — Kimi K3 triggered the question investors had been avoiding: A Chinese open-weight model helped touch off the worst week for chip stocks since April, as investors finally asked what $725 billion in AI capex is buying. The pattern is identical to the DeepSeek moment in January: a capable, cheap, open-weight Chinese model enters the market and immediately raises the question of whether frontier closed models — and the NVIDIA GPUs required to run them — are worth what the market is paying for them. The answer is more complicated than either the bulls or the bears want it to be. But the question is now back on the table for the second time in seven months.

3. The AI infrastructure battle is now physical — Bezos, Blackstone, and Samsung all made major moves in one week: Jeff Bezos personally invested in a new robotics actuator company that Blackstone also backed at roughly a $1 billion valuation — signalling that the next layer of AI infrastructure investment is moving into embodied AI hardware. Samsung confirmed its advanced packaging capacity for AI chips is sold out through Q2 2027. Competition in generative AI is no longer defined solely by benchmark results. Model providers must be able to operate reliable services at scale, control inference costs, and secure hardware capable of supporting sustained demand. The labs winning the infrastructure race are not necessarily the ones with the best models. They are the ones that can actually serve them.

A New Foundation for the Future of X

Map your AI infrastructure risk — before the next demand crash or supply squeeze finds it

Prompt: You are an AI infrastructure strategist. Three things happened this week that belong in every enterprise AI risk assessment: Kimi K3 crashed its own servers within 48 hours of launch due to demand — new subscriptions paused. Chip stocks had their worst week since April as investors questioned what $725 billion in AI capex is actually buying. Samsung's advanced packaging capacity for AI chips is now sold out through Q2 2027. The infrastructure layer beneath AI products is under more simultaneous pressure than at any point in the past two years.

Here is our current AI infrastructure setup: [describe your stack — which models you use, which cloud providers, how you handle capacity spikes, and what your current monthly AI spend is].

Help me map our infrastructure risk across four dimensions:

1. The demand spike scenario — if our primary AI model vendor paused new API access tomorrow due to capacity constraints — as Kimi K3 just did — what breaks first, how long does it take for our fallback to activate, and what is the customer-facing impact in the first two hours? If the answer is "we do not have a tested fallback," that is the finding.

2. The pricing shock scenario — chip stock volatility this week signals that the market is repricing AI infrastructure assumptions. If our primary model's pricing increased 50% next quarter — as Claude Sonnet 5 is increasing 50% from $2 to $3 per million input tokens on September 1 — what is the budget impact, which workloads become uneconomical, and what would we route differently?

3. The supply constraint scenario — Samsung's AI chip packaging is sold out through Q2 2027. New fab capacity does not come online until 2028. If GPU availability tightens significantly in the next six months, how does that affect our ability to self-host any models we depend on, and which of our cloud AI workloads are most exposed to capacity rationing?

4. The resilience scorecard — rate our current AI infrastructure on three dimensions: vendor diversity (1–5), geographic distribution (1–5), and fallback activation speed (1–5). For every dimension rated below 4, tell me the single most important action to take before Q3 ends.

End with a one-paragraph infrastructure posture statement I can share with our CTO — describing where we are exposed, what we are doing about it, and what the residual risk is after those actions.

AI news highlights

• Kimi K3 paused new subscriptions — demand crashed its own servers within 48 hours of launch — existing users unaffected. Full weights still July 27. The demand signal is the story: frontier-adjacent open models at competitive pricing have more buyers than anyone planned for.
• Chip stocks: worst week since April — $725B in AI capex, and investors are asking what they're buying — the DeepSeek pattern repeating. Capable, cheap, open Chinese model → question about whether closed frontier models justify their premium.
• SK Group Chair: customers want 60–100% more AI memory in 2027 — and no company has meaningful new capacity — foreign governments are treating AI memory access as economic security. The supply constraint is structural through at least 2028.
• Gemini 3.5 Pro: fourth delay, Flash stopgap under consideration — every week it is absent, frontier contracts get signed elsewhere. Google is the only major lab without a 2026 flagship in general production.
• Microsoft July patch fixed 570 vulnerabilities — with AI assistance — the largest AI-assisted security patch in history — AI is now part of the patch cycle itself. The same tools that create vulnerabilities are being used to find and fix them.
• DeepSeek July 24 migration deadline — three days left — deepseek-chat and deepseek-reasoner aliases error after cutover. Migrate today. Do not leave this for the weekend.
• Samsung advanced packaging for AI chips sold out through Q2 2027 — the memory and packaging layer beneath frontier AI is now as constrained as the logic layer. Infrastructure timelines are extending, not compressing.
• Claude Sonnet 5 introductory pricing ends August 31 — 41 days left at $2/$10 — standard $3/$15 from September 1. New tokenizer adds 1.0–1.35x token overhead. Recalibrate your cost models before the switch.

Trending AI tools

• Kimi K3 — 2.8T parameters, MIT license, 1M token vision context — full weights July 27, new subs paused
• Wispr Flow — Dictate in any app — writes in your voice, auto-edits, 100+ languages
• OpenCode — Open-source, model-agnostic coding agent — air-gapped, no vendor dependency
• Retrace — Debug AI agents by replaying and forking runs at the exact point of failure
• Granola — AI meeting notes on top of any call tool — no bot, no permissions needed
• Framer Agents — Design, write, and organise your site with agents — Claude Code and Codex compatible
• Claude Cowork Mobile — Delegate and monitor long agent sessions from iOS and Android
• Osaurus — Open-source agents that run 100% locally on your Mac — no cloud, no data exposure

That’s a Wrap

SPONSOR US

Get your business in front of over 90k+ AI professionals

8020AI is the world’s #1 AI Newsletter, Read by 90k+ professionals from leading companies such as Google, OpenAI, Meta, and Microsoft.

We've assisted in promoting Over 500 AI-Related Products. Will yours be the next?

What We Can Offer:

• Launch an Advertising Campaign

• Introduce New Product or Features

• Other Business Cooperation

Get in touch today.

Or Email our founder Alamin at [email protected]

FEEDBACK

How was your experience with 8020AI today?

Login or Subscribe to participate in polls.

If you have specific feedback or anything interesting you’d like to share, please let us know by replying to this email.