top of page

Q&A, Comments, Suggestions

Public·1 member

Chinese Nvdia?

https://www.wsj.com/tech/ai/alibaba-ai-chip-nvidia-f5dc96e3


  • Alibaba was long one of the biggest customers of American AI-chip leader Nvidia

  • Now it and other chip designers are filling the void left after Nvidia ran into regulatory barriersto selling its products in China.

  • Alibaba, founded by internet pioneer Jack Ma, is sometimes compared with Amazon.com because its biggest business is e-commerce, but it makes much of its money from the lower-profile business of cloud-computing services

6 Views
JA Soler
JA Soler
Sep 12

Helcio thank you for sharing this interesting article. My view is that Alibaba’s move to develop a new AI chip signals more than just another hardware milestone—it reflects a deep strategic shift. China has long been heavily dependent on foreign (especially U.S.) AI chip makers like Nvidia, but export restrictions and regulatory pressures have forced local companies to accelerate self‑reliance. Alibaba’s chip, designed for a broader range of inference tasks and built by a Chinese manufacturer, shows how necessity is driving innovation at pace.


What’s interesting is the dual tension this creates. On one hand, Alibaba wants compatibility with Nvidia’s platform so that engineers can reuse existing software, tools, and workflows. On the other hand, there are significant technical and capacity limitations: China’s domestic chip manufacturing remains less advanced in many respects, constrained by access to cutting‑edge semiconductor tech and production equipment. The new chip may not rival top U.S. chips in all respects, but its design is pragmatic—it covers inference, which is a huge part of existing AI workloads.


Strategically, this move strengthens China’s push toward an AI supply chain that is more autonomous. By filling the “void” left when Nvidia’s most powerful chips are blocked, Alibaba helps reduce vulnerability to external policy changes. The broader ecosystem—other chip designers, startups, and national funding—reinforces this trend: more investment, more local alternatives, more pressure to figure out how to do things at scale domestically.


But there are risks. Manufacturing capacity, heat dissipation, efficiency, and software/firmware integration remain challenges. Also, doing inference is one thing; training massive models (which require far more compute and hardware sophistication) is a different order. Alibaba’s chip seems focused on inference, which is important, but advancing local capability in training will likely be the tougher hurdle. Still, this is a meaningful step—it’s about reducing dependency, improving resilience, and shaping a future where geopolitical constraints force supply chain innovation.

Seizing the agentic AI advantage

🔗GenAI paradox: exploring AI use cases | McKinsey


The “Gen AI paradox”: nearly 80 % of firms use generative AI, yet most report little impact . McKinsey argues for shifting from reactive gen AI to proactive, agentic AI—equipped with autonomy, planning, memory, and integration. They introduce the “agentic AI mesh” architecture and stress that success requires addressing both technical and human challenges: trust, adoption, and governance. Organizational transformation must be driven by the CEO, with a move from siloed pilots to strategic programs, workforce upskilling, infrastructure adaptation, data productization, and agent-specific regulation.


The “Gen AI paradox”: the widespread adoption of generative AI has yielded minimal financial impact. This is because organizations deploy horizontal solutions—chatbots, copilots—without reimagining core processes . AI remains reactive; to unlock value, it must evolve to proactive, goal-driven systems.


“Agentic AI mesh" implies systems that plan, act autonomously, remember, and integrate across workflows . It distinguishes single-agent from…


6 Views

JPM on AI Security (interesting LinkedIn Post)

 JP Morgan just released an Open Letter to their third party suppliers, and sounded the AI security alarm! 


The financial giant sees what others are missing:

Companies rushed to deploy AI without understanding the consequences. The mandate was clear: innovate or die. But JP Morgan's latest security assessment reveals that: 

• 78% of enterprise AI deployments lack proper security protocols

• Most companies can't explain how their AI makes decisions

• Security vulnerabilities have increased 3x since mass AI adoption


7 Views
JA Soler
JA Soler
May 01

Thank you Carlos, interesting post. JP Morgan's proactive stance underscores a crucial reality: AI security can't be an afterthought. The figures are alarming, especially for sectors handling sensitive data. Companies need to balance innovation with robust risk management, and JPM's recommendations offer a strong starting point. Curious to see how other financial institutions respond to this wake-up call!

Can AI Crack the Code on AT1 Bond Calls?

Following Deutsche Bank's decision to not call one of its Additional Tier 1 ($1.25 billion 4.789%, which will reset to a coupon of approximately 8.46% and a new call date in 2030), I wanted to open up the forum regarding the potential use of AI in predicting AT1 call decisions with precision.


A key factor in DB’s decision was FX losses. Calling and replacing both bonds would have resulted in a €400 million FX loss due to changes in exchange rates since the bonds were issued in USD in 2014. This loss would have been significant relative to DB’s profitability targets, particularly as the bank aims for a 10% return on tangible equity (RoTE) by 2025, well below the 13% average seen by European peers.


Banks weigh multiple complex and often unpredictable factors when deciding whether to call AT1 bonds, from refinancing costs to investor sentiment and FX losses. While…


26 Views
bottom of page