A UNCTAD report reveals that AI profits concentrate at the top of the supply chain, with developing nations seeing minimal returns. For the Nvidia GB200 server, raw materials accounted for just 0.2% of the final sales price, while physical production workers earned only 4.3% of traceable value. The report calls on developing countries to build local supplier networks and skilled workforces to capture more value.
Vitalik Buterin highlighted a key difference between Bitcoin mining and AI inference costs. Bitcoin mining splits expenses nearly evenly between hardware and electricity, while AI costs lean heavily on hardware with much smaller energy bills. This gives AI operators an advantage in competing for power plant access, especially as electricity prices rise. Bitcoin mining currently runs about $93,700 per coin.
Anthropic's internal reports show a sharp gap between its staff and the general public on AI's future. By August 2026, AI agents at the company outnumbered employees six to one, and productivity climbed as agents worked more hours. Ninety-one percent of Anthropic staff expect Nobel-level discoveries by 2030, compared to just 39% of US adults. Over half of staff believe AI could handle 80% of their current work by 2030, but only 5% of Americans share that view.
On the security front, Rubrik launched Code Guardian, an AI-powered code vulnerability scanner built with Anthropic's Mythos model. The tool checks code for security flaws in isolated air-gapped environments. This comes after a CERT-In warning about rogue AI agents accessing live systems in Indian organizations. Meanwhile, experts stress that agentic AI security needs multiple defense layers following the Hugging Face incident that exposed agent vulnerabilities. Only 27% of leaders feel confident their organizations track all AI systems in use.
The digital legacy industry is growing rapidly as "deadbots" — AI chatbots modeled on deceased individuals — raise ethical and emotional concerns. The sector was valued at $22 billion in 2024 and could hit $80 billion by 2034. Researchers found that grieving families may face added emotional burdens and commercial pressures from these tools. Separately, Hong Kong medical student Claudia Chan Shu-ying is developing CancerCompanion, an AI tool offering emotional support to cancer patients between appointments, with safeguards to route medical questions to healthcare teams and flag self-harm risks.
Key Takeaways
- UNCTAD reports that workers in physical AI production earn just 4.3% of traceable value, while raw materials make up only 0.2% of the Nvidia GB200 server's price
- Vitalik Buterin notes AI inference costs are mostly hardware, giving AI operators an edge over Bitcoin miners in competing for power plant access
- Anthropic's AI agents outnumbered employees six to one by August 2026, with productivity gains as agents worked more hours
- 91% of Anthropic staff expect Nobel-level AI discoveries by 2030, compared to 39% of US adults
- Rubrik launched Code Guardian, an AI code vulnerability scanner built using Anthropic's Mythos model
- Only 27% of leaders are confident their organizations track all AI systems in use after the Hugging Face agent vulnerability incident
- The digital legacy industry was valued at $22 billion in 2024 and could reach $80 billion by 2034
- Canada aims to recruit and train 80,000 to 100,000 Red Seal trades workers by 2030-31 to address skills shortages tied to major projects
- CancerCompanion, an AI tool by Hong Kong medical student Claudia Chan, provides emotional support to cancer patients and flags self-harm risks for human review
- Canada's Team Canada Strong plan and Future Skills Centre recommend skills-first hiring and AI tools to help small firms train workers
AI profits flow to top of supply chain while developing nations see little gain
A UNCTAD report shows that most AI profits go to firms that design technology and own intellectual property. Developing countries supply raw materials and assemble parts but earn very little. For the Nvidia GB200 server, raw materials made up only 0.2% of the final sales price. Workers involved in physical production earned just 4.3% of the traceable value. The report urges developing countries to build local suppliers and skilled workforces.
Vitalik Buterin explains why AI could outcompete Bitcoin miners on energy costs
Vitalik Buterin said Bitcoin mining costs are split nearly evenly between hardware and electricity. AI inference costs are mostly hardware with much smaller electricity bills. This means AI operators can handle higher electricity prices better than Bitcoin miners. Bitcoin mining costs total about $93,700 per coin according to his data. The difference in energy use may determine who wins access to power plants in the future.
AI platforms transform how manufacturers connect with buyers
AI is reshaping the manufacturing marketplace by connecting buyers with contract manufacturers through digital platforms. Xometry uses AI to give customers instant quotes and route orders to vetted shops. Job shops face high variability in demand and only about a 30% win rate on quotes. New customer revenue averaged just 10% of total revenue in 2024. Online platforms help shops quote faster and reduce unused capacity.
Anthropic report shows staff and public differ on AI future expectations
Anthropic published two reports on AI's impact on work and expectations. By August 2026, AI agents outnumbered employees six to one at the company. Agents worked more hours than employees by March 2026 and productivity rose. 91% of Anthropic staff expect AI to make Nobel-level discoveries by 2030 compared to 39% of US adults. More than half of staff believe AI could do 80% of their current work by 2030. Only 5% of Americans share that expectation.
Canada pushes employers to train workers as major projects speed ahead
Canada is fast-tracking major projects like ports and data centers but faces a skills shortage. MPs told employers they must take on more training responsibility. Ottawa's Team Canada Strong plan aims to recruit and train 80,000 to 100,000 Red Seal trades workers by 2030-31. The Future Skills Centre recommends models like skills-first hiring and AI for small firms. A proctored exam can verify skills for employers seeking training partners.
Rubrik launches AI tool to secure code from rogue agents
Rubrik unveiled Code Guardian, an AI-powered code vulnerability scanner built using Anthropic's Mythos. The tool scans code for security flaws in isolated air-gapped environments. This launch follows a CERT-In warning about rogue AI agents accessing Indian organizations' live systems. Rubrik executives discussed the need for flexible security approaches in a rapidly evolving AI landscape.
AI deadbots raise difficult questions about grief and ethics
Researchers around the world are studying "deadbots," AI chatbots modeled on deceased people using their texts and voicemails. A global study examines how these digital simulations affect the grieving process. The digital legacy industry was valued at $22 billion in 2024 and could reach $80 billion by 2034. One researcher found that grieving families may face emotional burdens and commercial pressures from these tools.
Agentic AI security needs multiple layers of protection
Experts say AI agent security requires layered governance because no single control can prevent all failures. After the Hugging Face incident exposed agent vulnerabilities, the industry is moving toward independent layers of defense. Kyndryl recommends four steps: inventory all agents, enforce high-risk rules in code, add independent monitoring layers, and collect evidence from day one. Only 27% of leaders are confident their organizations track all AI systems in use.
Hong Kong student builds AI companion for cancer patients
Claudia Chan Shu-ying, a medical student at the Chinese University of Hong Kong, is developing CancerCompanion. The AI tool provides emotional support to cancer patients between appointments, including at night when help is not readily available. It directs medical questions to healthcare teams and flags potential self-harm cases for human review. Chan is a finalist in the youth category of the Spirit of Hong Kong Awards.
Sources
- AI-driven profits concentrated at top despite widespread global AI adoption
- Vitalik Buterin Reveals Why AI Could Push Bitcoin Miners Out
- AI dives deep into the manufacturing marketplace
- Working at the frontier shapes expectations for the future
- Ottawa Fast-Tracks Major Projects While MPs Tell Employers to Train More. What It Means for AI Training Partners
- ‘Build for flexibility’: Rubrik executives on the race to secure code amid ‘rogue’ AI agents
- AI can help us talk with the dead. Should we?
- Why agentic AI security requires layered governance
- A companion between appointments: Hong Kong student’s AI tool for cancer patients
Comments
Please log in to post a comment.