PLUS: China’s EUV breakthrough, Salesforce AI agents, Gen Z fitness adoption

Morning all, The AI memory market is in crisis, with prices skyrocketing 55% this quarter as chipmakers scramble to meet booming demand for AI infrastructure. Major suppliers like Samsung and Micron are shifting nearly all production away from consumer devices, signaling a seismic shift in global tech priorities.

This shortage exposes a critical bottleneck - GPUs now spend over 40% of their time idle while waiting for data. With new memory solutions years away, can the industry build sustainable infrastructure fast enough to keep pace with AI's exponential growth?

Today's dots:

• AI memory crunch drives prices up 55%
• China's EUV breakthrough despite US sanctions
• Salesforce’s AI agents transform marketing emails
• Gen Z drives India’s fitness revolution
• ChatGPT Health enters medical space

AI Memory Crunch Slams Brakes on Innovation

Here's the thing: Surging demand for AI chips has triggered an unprecedented RAM shortage, pushing prices up 55% this quarter and forcing memory giants like Micron and Samsung to abandon consumer markets.

Let's unpack that:

• Making specialised HBM memory for AI chips like Nvidia's Rubin consumes 3X more production capacity than standard RAM - essentially prioritising one AI server over 3,000 smartphones' worth of memory.
• Suppliers like Samsung are redirecting 90%+ of output towards big tech clients, with Micron already halting consumer PC RAM sales entirely to meet AI demand.
• Consumer tech margins face pressure as memory now eats 20% of device costs (up from 10% last year) - Dell warns price hikes are "inevitable", challenging Apple's claims of minimal impact.
• Researchers call this the "memory wall" - where GPUs sit idle 40%+ of the time waiting for data, stalling the very AI acceleration we need to overcome bottlenecks.
• Breakthroughs like Majestic Labs' 128TB memory systems hint at future workarounds, but new Micron factories won't come online until 2027 at earliest.

If you remember nothing else: This crunch proves AI's infrastructure demands now dictate global tech economics. While painful short-term, it's accelerating investment in memory innovation that'll ultimately make smarter AI accessible beyond Silicon Valley giants.

China’s AI Climb: Innovation vs. Silicon Limits

Here's the thing: Chinese AI researchers have cracked prototype EUV chipmaking tech and raised billions through IPOs despite US sanctions - creating a new innovation playbook that could redefine global AI competition (Reuters exclusive).

Let's unpack that:

• Beijing scientists completed a working EUV lithography prototype - the holy grail of advanced chipmaking - though production-grade chips remain years away
• Material advantages: China's cheap electricity and infrastructure outpaces rivals, but lacks ASML-grade lithography machines and mature AI software ecosystems
• Investment gap persists (US spends 10-100× more on compute), forcing breakthrough efficiency hacks like Alibaba's algorithm-hardware co-design that runs LLMs on budget chips
• IPO boom: Zhipu AI's recent HK$4.35B listing signals market confidence in homegrown alternatives to US tech
• Cultural shift: Younger founders embrace Silicon Valley-style risk-taking - previously uncommon in China's corporate landscape

If you remember nothing else: Western sanctions forced China to innovate differently, not slowly. Expect faster cycle times on hardware breakthroughs as capital floods semiconductor moonshots - with 2027-2030 as the make-or-break window.

Salesforce's AI Agents Turn Marketing Emails Into Revenue Streams

Here's the thing: Salesforce's new Agentforce Commerce tools turn static email/SMS campaigns into AI-driven sales conversations - personalising offers and closing deals directly in your customer's inbox.

Let's unpack that:

• Two-way SMS/email features let customers reply directly to marketing messages, with AI agents handling up to 10,000+ responses per campaign while maintaining brand compliance
• Personalised deal-making: Agents dynamically adjust offers based on customer responses - like extending coupon validity or bundling products when shoppers ask 'Can I use this with last week's promo?'
• Backed by new Contextual Search (currently in beta) that analyses conversation history to serve hyper-relevant product recommendations
• Forces marketers to rethink workflows: What happens when 7% of your mailing list suddenly becomes active negotiators instead of passive clickers?
• Already live for select customers after being showcased at January's NRF '26 event, with full rollout through February

If you remember nothing else: Retailers now have AI tools to directly convert customer engagement into closed sales. This moves marketing tech from 'broadcast mode' to personalised deal-closing engines that work while your team sleeps.

Gen Z Fuels India's AI Fitness Surge

Here's the thing: Young Indians aged 18-29 now make up 75% of AI fitness tool users on Boat Crest's platform - proving AI coaches aren't just gadgets, but behaviour-changing companions. (Boat Crest 2025 Report)

Let's unpack that:

• Gen Z users log 35% more workout minutes with AI coaches than other groups, tracking everything from cricket matches to treadmill sessions
• The platform saw 1 trillion steps logged last year - enough to walk around Earth 18 times - with walking dominating 55% of activities
• Wednesdays unexpectedly became peak fitness days, overtaking weekends as users embrace mid-week health boosts
• 20,000-step challenges proved most popular - showing how AI-powered nudges help build consistent habits better than extreme goals
• Boat's CPO calls this "where wearables graduate from trackers to lifestyle partners" through real-time adaptation

If you remember nothing else: This isn't just about counting steps. It's how AI coaches make fitness feel accessible through personalised, bite-sized goals. Expect this behavioural nudge tech to spread into workplace wellness and education next.

ChatGPT Health: Your AI-Powered Medical Sidekick

Here's the thing: OpenAI just launched ChatGPT Health - a HIPAA-compliant medical assistant that analyses your health records and wearable data to prepare you for doctor visits, marking AI's biggest step yet into regulated healthcare.

Let's unpack that:

• You can securely connect medical records and apps like Apple Health to get personalised interpretations of bloodwork and health trends, acting like a 24/7 medical translator for your data
• Built with input from 260+ doctors who've reviewed 600k+ responses, it specialises in explaining complex terms in plain language - perfect for decoding those confusing lab results
• Enhanced security keeps health chats encrypted and isolated from regular conversations, with options to instantly delete sensitive data if needed
• Early adopters are using it to spot patterns between sleep stats and blood pressure, then generate tailored questions for their GP during 15-minute appointments
• Currently US-only for medical integrations, but all users can join the waitlist as global rollout begins

If you remember nothing else: This won't replace your doctor, but it might finally help you understand what they're saying about your cholesterol levels. AI's healthcare debut shows how technology can make complex systems navigable - starting with our own bodies.

The Shortlist

Wharton argues successful enterprise AI implementation requires triple the expected costs and workforce retraining, not headcount reductions. Professor Peter Cappelli reveals most 'AI pilots' fail to match consultant hype despite case studies showing marginal efficiency gains.

John Nosta claims generative AI promotes 'anti-intelligence' by prioritising fluency over understanding in workplace cognition. The innovation theorist warns polished outputs create dependency while erasing critical thinking's 'productive friction'.

Physical AI dominated CES 2026 as embodied systems moved beyond chatbots. NVIDIA's robotics platforms and Hyundai's award-winning Atlas humanoid signaled industry readiness for real-world deployment, though experts note high simulation/testing costs remain barriers.

Wall Street bet $900B+ on enterprise software stocks as analysts dismissed bubble fears. D.A. Davidson's report highlights Commvault and Datadog as key infrastructure plays for managing AI's 'real-world implementation complexity'.