PLUS: The coming AI 'memory tax' on new devices and the disruption facing software developers
Good morning
AI’s shift from experimentation to real revenue just hit a new extreme. An agent-first startup has set a speed record that would have sounded implausible a year ago, while the infrastructure beneath the AI boom is starting to leak costs directly into consumer hardware. At the same time, the pace of adoption is compressing entire industry transitions into just a few years.
Today’s stories all point to the same underlying theme. AI is no longer waiting for permission. It is scaling faster than organizations, supply chains, and career paths were built to absorb, forcing hard adjustments in how products are priced, built, and staffed.
In today’s Next in AI:
Manus becomes the fastest company ever to reach $100M ARR
Why AI workloads are creating a hidden “memory tax” on new devices
How AI is compressing decades of developer disruption into a few years
Scribe’s bet that workflow visibility is the missing layer for enterprise AI
Manus hits $100M in record time
Next in AI: AI agent startup Manus just became the fastest company to hit $100M in annual recurring revenue, reaching the milestone in only eight months after its public launch.
Decoded:
Behind the revenue is immense scale: Manus has already processed over 147T tokens and created more than 80M virtual computers for its users.
The company's growth accelerated to over 20% month-over-month following the release of Manus 1.5, directly linking product updates to its rapid market adoption.
Beyond its product, the company is shaping the field by publishing guides on topics like Context Engineering, which have become an industry standard.
Why It Matters: Manus's hyper-growth validates the massive commercial appetite for AI that moves beyond chat to actively delegate and automate complex work. This milestone sets a new benchmark for AI startups, signaling a major shift toward practical, agent-based computing.
The AI Memory Tax
Next in AI: The massive demand for memory chips from AI data centers is creating a global shortage. Analysts predict the cost will soon hit your wallet when you buy new PCs, phones, and other devices.
Decoded:
The demand isn't just a temporary blip because AI workloads are fundamentally built around memory, requiring huge, persistent footprints and extreme bandwidth to function properly.
Major industry players are confirming the trend, with chipmaker Micron forecasting a prolonged shortage and Dell Technologies noted the rising costs will likely be passed on to customers.
There is no short-term fix in sight, as analysts project that chip makers will max out their current production capacity by the end of 2026, with the next major factory not expected to be operational until 2027.
Why It Matters: The AI boom is now directly impacting the consumer tech supply chain, meaning the price of your next upgrade is likely going up. This 'memory tax' shows how foundational AI infrastructure demands are starting to reshape the broader hardware market for everyone.
AI's Disruption: No Ten-Year Warning for Devs
Next in AI: A new analysis argues that software developers don't have a decade to adapt to AI—the industry is compressing the internet's disruption of travel agents into a few short years, driven by rapidly improving agent capabilities tracked by organizations like METR.
Decoded:
The adoption curve is nearly vertical. While less than 5% of travel was booked online by 1999, LLMs are already used by over 40% of the entire US population, and the rate among developers has hit an incredible 84% in just three years.
There are clear economic parallels to the past. Just as airlines dramatically cut commissions in 1995, weakening travel agencies before the internet boom, today’s pullback in VC funding is pressuring the software job market as AI's impact grows.
Survival means moving up the value chain. The developers who thrive will be those who focus on domain knowledge and end-to-end ownership, using AI to bridge skill gaps and solve entire business problems, not just translate requirements into code.
Why It Matters: Like the generalist travel agents of the past, developers focused on the most commoditized coding tasks face the biggest risk. This is the moment to lean in and use AI as a lever to broaden your skills across the entire stack, making you more valuable than ever.
Scribe's Billion-Dollar Blueprint
Next in AI: Workflow AI startup Scribe just secured $75 million at a $1.3 billion valuation to solve a critical problem for enterprises: giving AI the context it needs to actually work.
Decoded:
Scribe's platform records how expert employees complete tasks, automatically generating step-by-step guides to create a map of company processes. Its newer Optimize tool then analyzes this data to pinpoint the best opportunities for automation.
The company’s core thesis is that “AI can’t improve what it can’t see.” By first documenting how work actually gets done, Scribe provides the essential context that most AI implementations lack, preventing them from failing.
Scribe already boasts impressive traction with over 75,000 customers, including major players like LinkedIn and T-Mobile, and counts 44% of the Fortune 500 as paying clients.
Why It Matters: Many companies are struggling to get real value from AI because their internal processes are a black box. Scribe is betting big that documenting and understanding workflows is the foundational layer required before any successful AI strategy can be built.
AI Pulse
NVIDIA faces growing investor scrutiny over its complex, circular financing deals with major customers and opaque multi-billion dollar partnerships with governments like South Korea.
US Senators proposed the GUARD Act to protect minors from AI companions, as Senator Bernie Sanders separately called for a moratorium on new data centers over fears of mass job displacement.
A new framework distinguishes between "Job" tasks to be automated by AI and "Gym" tasks that maintain critical thinking skills, urging users to be deliberate about which work they delegate.
Developers debate the creation of "anti-AI" open-source licenses as a German court rules that training models on copyrighted work can constitute a copyright violation.