AI affordability wakeup call, Anthropic's $65bn mega round, and India's first 12nm AI chip -- Weekly AI roundup

From Anthropic's $65 billion mega funding round to the announcement of India's first homegrown 12nm AI chip, the world of AI has seen a lot of action this week. AI defined the force of global finance, cybersecurity, and corporate strategy. In the week that went by, we saw Anthropic raising its value to $965 billion while China's DeepSeek decided to cut prices by 75 per cent. Big Tech realised that AI infrastructure isn't cheap to maintain and hence, ordered scaling down on encouraging AI usage, while Nvidia's Jensen Huang, in the midst of global political disruption, is joining the Tsinghua Advisory Board.

What matters more, however, is that these weren't isolated events on their own. All these were interconnected signals that AI is now getting ready for a higher-level game, forming an infrastructure layer that supports a lot of institutions and their infrastructure.

Hence, without further ado, let's take a quick look at everything the AI world had to throw at us.

1. Anthropic now the world's most valuable AI firm

Anthropic's overtook OpenAI to become the most valuable AI company yet after the latest $65 billion funding round - something that shall dominate business pages for months. The AI company, led by Dario Amodei, closed a Series H round at a post-money valuation of $965 billion, making it Silicon Valley's most valuable AI startup. Led by Altimeter Capital, this $65 billion round is reportedly the AI startup's final private fundraise before a highly anticipated initial public offering (IPO).

At nearly $1 trillion, Anthropic is now worth more than major Fortune 500 companies combined despite being a private company. This valuation reflects investor confidence despite its rivals making great strides in the AI space over the week.

2. Big Tech hits the AI cost wall

The AI affordability crisis hits Meta, Amazon, and Uber, emerging from token-based pricing models that have driven operational costs higher than the returns on investment. AI companies, which once viewed using AI as a cost-saving effort, have found enterprise-grade generative AI to be remarkably expensive, owing to the massive token consumption of autonomous agents and code generation. This has forced a shift to strict budget control. Uber exhausted its entire 2026 budget for internal AI coding tools in just four months, leading to a reevaluation of agentic software engineering.

Similarly, Amazon moved away from its initial phase of tokenmaxxing toward rigid cost controls, prioritising practical business outcomes. Meta's rising infrastructure expenses have forced teams to implement internal benchmarking frameworks like 'Claudeonomics' to strictly audit departmental resource consumption.

3. Anthropic also released Claude 4.8

On the sidelines, the company released Claude Opus 4.8 - an upgraded model that delivers notable performance improvements across coding, reasoning, and agentic workflows. A key highlight of this update is enhanced model reliability and honesty, with Opus 4.8 being four times less likely to overlook flaws in its own code. It is also significantly better at proactively flagging uncertainties rather than making unsupported claims.

Alongside the model, Anthropic introduced many major features to other Claide products, including 'dynamic workflows' in Claude Code. This allows the system to run hundreds of parallel subagents to tackle massive, codebase-scale migrations, and an 'effort control' setting that lets users manually adjust how deeply Claude reasons through a task on claude.ai.

Anthropic's play on price also took everyone's note, with regular API pricing remaining identical to Opus 4.7, and fast mode now being three times cheaper. Anthropic also teased that its Mythos-class models are expected to release in the coming weeks.

Since we touched upon price...

4. DeepSeek's permanent 75% price cut

In a move that sent shockwaves through the AI infrastructure market, DeepSeek made its 75% price discount on the flagship V4-Pro model a permanent one. The 1.6-trillion parameter model now costs just $0.83 per million tokens for output, down from $3.48 -- a 25% drop from the original API cost.

What was supposed to be a promotional discount ending May 31 is now the new baseline. AI analysts are calling it the beginning of the AI pricing war, where DeepSeek has taken the lead. This permanent price cut will accelerate the shift from chatbots to autonomous systems that can perform complex workflows.

5. Project Lightwell: IBM's $5 billion security counterstrike

While Anthropic's cybersecurity model raised alarms, IBM and Red Hat responded with a counter-initiative. The companies announced Project Lightwell - a $5 billion commitment to deploy 20,000 engineers supported by advanced AI to establish a trusted enterprise protector for open-source security. This represents the largest coordinated effort to secure software supply chains in history, acknowledging that AI-powered attacks are becoming the primary threat to digital infrastructure. This ends up being crucial, as AI models become more capable at discovering vulnerabilities faster than humans can fix them - IBM's tool can weaponise those discoveries at an unprecedented scale.

6. IBM India head urges upskilling millions

IBM India and South Asia Managing Director Sandip Patel has warned that Indian IT companies must urgently reskill millions of workers to remain competitive as AI rapidly transforms software development, coding, and business operations. Highlighting that only about 30% of India's current workforce is AI literate, Patel stressed that rapid, large-scale upskilling could leverage the country's vast youth demographic to establish India as the "skill capital" of the world for AI by 2030.

7. Faith meets tech: Pope and Anthropic partner on AI ethics

This week, Pope Leo XIV made history by personally presenting his first papal encyclical, Magnifica Humanitas: On safeguarding the human person in the time of artificial intelligence, alongside Anthropic co-founder Christopher Olah. The landmark event at the Vatican brought together the head of the Catholic Church and the self-described atheist tech leader to advocate for a rare partnership between religious institutions and the tech industry.

In his encyclical, Pope Leo XIV warned against the concentration of technological power and called for the global "disarming" of the AI arms race to protect against algorithmic bias, automated warfare, and mass labour displacement. Olah strongly echoed these moral concerns, candidly admitting that frontier labs face immense commercial and geopolitical pressures that can conflict with safety. He emphasised that the tech sector urgently needs independent, outside moral voices, like the Church and civil society, to provide rigorous oversight and ensure the technology is steered toward the common good.

8. Jensen Huang joins Tsinghua Advisory Board

The week also saw Jensen Huang, Nvidia's CEO, join Tsinghua University's advisory board alongside Tim Cook and Elon Musk, accepting a seat on the school of economics and management. The move comes amid active US chip export controls on China, making Huang's role at Beijing's most prestigious university significant geopolitically.

On the contrary, the Chinese government has placed strict overseas travel restrictions on top AI scientists working at private tech powerhouses like DeepSeek and Alibaba in a bid to protect domestic IP and halt brain drain to US labs.

9. AGI possible by 2029, says DeepMind's Demis Hassabis

In a recent interview with Axios, Google DeepMind CEO Demis Hassabis turned heads by tightening his timeline for Artificial General Intelligence (AGI). Hassabis declared its arrival by 2029 as a "real possibility." This marks a sharp contraction from his previous, more conservative 2030 horizon. Hassabis admitted that his dramatic "singularity" framing was an intentional policy pressure tool designed to shock sluggish governments and economists into accelerating their regulatory preparation.

10. Zoho-backed Netrasemi unveils India's first homegrown 12nm AI chip

Kerala-based semiconductor startup Netrasemi has officially launched its maiden high-end 12-nanometer AI system-on-chip (SoC), named the A2000, with commercial mass production scheduled to begin by the end of 2026. Backed by Rs 125 crore in funding from prominent investors like Zoho and Unicorn India Ventures, as well as support from the Indian government's Design Linked Incentive (DLI) scheme, the A2000 has successfully cleared initial silicon bring-up testing. This chipset integrates a proprietary neural processing unit (NPU), video processing unit (VPU), and image signal processor (ISP) to execute high-performance inference workloads locally on devices without cloud reliance. The chip is suited for advanced edge AI applications like smart surveillance cameras, robotics, and video gateways.