PLUS: a critical flaw hits AI agents, an AI jobs reality check, and the Pope's take on ethics
Happy reading
The server CPU market has long been dominated by Intel and AMD, but a powerful new challenger has emerged. NVIDIA is entering the fray with Vera, a chip designed specifically for the growing demands of agentic AI.
With systems already shipping to top AI companies, the move signals a major shift toward specialized hardware for high-value AI tasks. Does this mark the beginning of the end for the general-purpose chip's reign in the data center?
In today's Next in AI:
NVIDIA's new CPU for agentic AI
Critical 'BadHost' flaw hits AI agents
A reality check on AI's impact on jobs
The Pope releases an ethical framework for AI
NVIDIA's New Agentic CPU
Next in AI: NVIDIA is making a major play in the data center with its latest CPU, Vera, a chip purpose-built for the demands of agentic AI. Early results show it outperforming top competitors in key workloads.
Explained:
Vera delivered a 1.5x performance advantage over a new 128-core x86 processor in a series of independent benchmarks covering tasks like code compilation, data compression, and database management.
The chip’s design centers on memory efficiency, using LPDDR5X memory to achieve a massive 1.2 TB/s of bandwidth at a fraction of the power of traditional server memory, which is critical for running many AI tasks in parallel.
This isn't just a prototype; NVIDIA is already delivering systems to major AI companies and cloud providers, signaling a confident and rapid move into the market.
Why It Matters: NVIDIA's entry creates the most formidable challenge yet to Intel and AMD's long-standing dominance in the server CPU market. This signals a new era of specialized hardware where chips are increasingly designed to accelerate specific, high-value AI workloads beyond just training models.
Critical Flaw Hits AI Agents
Next in AI: A critical vulnerability dubbed 'BadHost' has been discovered in a widely-used Python framework, exposing millions of AI agents and tools to potential data theft and server takeover. A patch has been released to address the issue.
Explained:
The flaw, tracked as CVE-2026-48710, exists in Starlette, an open-source framework that underpins popular AI development tools like FastAPI, vLLM, and LiteLLM.
It allows attackers to bypass path-based authorization by sending a specially crafted HTTP Host header, granting them unauthorized access to sensitive server resources.
Security researchers warn the exploit is trivial to execute and could expose everything from clinical trial databases and user PII to private source code and cloud credentials. An online scanner is available to check for exposure.
Why It Matters: This vulnerability highlights the significant supply chain risks embedded in the rapidly expanding AI ecosystem. It serves as a critical reminder that the security of foundational open-source tools is paramount for protecting the next generation of AI applications.
AI's $350B Economic Engine
Next Take: Big Tech giants are on track to spend over $350 billion on AI infrastructure this year, an investment so massive that economists see it as a powerful boost to an otherwise decelerating economy.
Decoded:
The $350 billion figure represents the combined annual spending from tech titans Google, Meta, Amazon, and Microsoft alone.
This capital influx is primarily directed at building and equipping sprawling AI data centers, the physical backbone required to power modern AI.
Economists and analysts suggest this spending spree acts as a significant countervailing force, potentially offsetting downward trends in other sectors of the economy.
Why it Matters: This massive infrastructure investment shows that the AI boom is not just digital but a physical, economic force. Such foundational spending signals a long-term commitment from Big Tech…
The Pope on AI Ethics
Next in AI: The Vatican has released a detailed encyclical, Magnifica Humanitas, offering a comprehensive ethical framework for AI development and deployment. It calls for prioritizing human dignity and accountability in an era of rapid technological advancement.
Explained:
The document emphasizes human accountability, warning that critical decisions affecting lives in hiring, credit, or justice should not be fully delegated to automated systems that lack compassion and cannot be held responsible.
It raises concerns about AI amplifying the power of a few, arguing that the technology could worsen inequality if not guided by principles of social justice and solidarity. The encyclical also suggests data should be treated as a common good rather than a privately owned commodity.
Pope Leo XIV also highlights the interpretability problem, noting that current AI systems are often cultivated rather than explicitly built. This opacity makes it difficult to understand their reasoning, assign responsibility for errors, and correct for cultural biases.
Why It Matters: This marks a significant moment where one of the world's oldest institutions is directly addressing the core ethical challenges of modern AI. The encyclical provides a moral counterpoint to the tech industry's rapid innovation, urging developers and policymakers to build a future where technology serves the common good.
AI Pulse
Uber revealed it burned through its entire 2026 AI coding tools budget in just four months, with its COO now questioning the link between high tool usage and shipping more useful consumer features.
Stack Overflow is pivoting its business model as its public Q&A forum traffic plummets due to AI, now licensing its data to AI companies and selling enterprise AI solutions.
Spotify defended its new AI remix feature as a "controlled" alternative to unregulated AI music, allowing users to create remixes from a catalog of consenting Universal Music Group artists.
US law enforcement is reportedly classifying "anti-technology extremism" as an emerging domestic threat, with internal reports showing surveillance of groups and activities critical of AI and data center proliferation.