Microsoft Build 2025 Developer Conference opened grandly on 5/19, where CEO Satya Nadella engaged in conversations with NVIDIA founder Jensen Huang, OpenAI CEO Sam Altman, and xAI founder Musk. They simultaneously revealed a surprising reality: "AI has officially transitioned from laboratories into daily life and workplaces, capable of helping with overtime programming, providing customer service for businesses, while adhering to ethical standards and avoiding nonsensical statements."
AI Factory Replaces Traditional Data Centers, Jensen Huang: The Strongest Black Technology is AI Computers
Jensen Huang first stated that two years ago, they collaborated with Microsoft to build the world's largest AI supercomputer on the Azure cloud platform. Now, the new generation Grace Blackwell (GB200) with performance surging 40-fold has been officially mass-produced, symbolizing AI supercomputers as the current most powerful "black technology".
Software Growth Line Profit Key, Old Chips Can Unleash New Capabilities Through Software
Jensen Huang emphasized that even older chips like the Hopper GPU can significantly improve computational efficiency through software updates. He introduced techniques like in-flight batching and speculative decoding that can accelerate AI processing without hardware replacement.
Codex Agent Launches AI Collaboration Era, Sam Altman: AI Can Also Take Independent Projects
Altman announced the new Codex Agent, which can independently fix bugs, build features, explain code, and handle multiple tasks simultaneously. He directly stated: "AI is not just an assistant, but a capable colleague who can take independent projects."
Discussing 30 Years of Development Tool Evolution, from CLI, Copilot to Codex Agent
Altman indicated that future model development will focus on three aspects:
- Easier to use, capable of automatically determining optimal solutions
- Supporting multi-modal inputs like text, voice, and images
- Deeply integrating with tools to make development processes smoother
AI development is progressing very quickly, with the latest products from a year ago potentially becoming obsolete within six months. Altman also called out to developers: "The earlier you get involved, the greater the returns."
Musk announced during an interview that Grok will be upgraded to version 3.5, with the previous design concept aimed at making AI reasoning more similar to a "physicist's thinking", starting from the most fundamental natural laws and gradually constructing logic.
He stated directly: "AI's reasoning and judgment cannot deviate from the basic logic of the physical world. If it violates this, it is wrong."
Musk emphasized that for AI to operate smoothly in the real world, it must have the ability to align with reality. This means that whether it's self-driving cars or humanoid robots, they only have practical value when they exhibit reasonable behavior while adhering to physical rules.
Currently, Grok has been applied to Tesla and SpaceX's customer service systems, described by colleagues as an AI colleague that never complains and is very polite, significantly improving customer satisfaction and processing efficiency.
Musk revealed that Grok's next step is to integrate language, visual, and action capabilities, and collaborate with other AIs to form a Multi-Agent System, with future applications covering:
- Tesla's self-driving system
- Optimus humanoid robot
- SpaceX space mission assistance system
Musk finally urged developers to use Grok more and provide feedback, emphasizing that community feedback is the driving force for Grok's growth.
From Huang Renxun's AI factory to Altman's Codex agent to Musk's physical reasoning model, these three discussions depict a shared vision:
"AI will no longer be just an auxiliary tool, but become a collaborative, deployable, and executable entity platform."
Microsoft is not just providing tools, but building a complete system from models, training, deployment, integration, governance to an open ecosystem, allowing developers to shift from writing code to creating AI agents and truly participate in the next AI golden decade.
ARK Invest founder Cathie Wood, in an interview this morning 5/20, discussed the current US economy, AI, and biotech innovation, also warning that traditional industries that do not embrace AI technology may soon be eliminated. She is particularly optimistic about tech stocks like Palantir, Coinbase, and Roku, but also mentioned: "The current market is too focused on AI infrastructure, but industries that actually use AI for innovative applications like biotech and healthcare are the real dark horses."
Wood first stated that although the 4/2 tariff liberation day initially caused market confusion, with Treasury Secretary Scott Bessant taking over negotiations and non-tariff barriers gradually being relaxed, market sentiment has begun to turn.
- Food safety standards and hormone restrictions for US beef
Carbon emission regulations and production certification for US ethanol
Quota, administrative review, and technical standard restrictions
If these restrictions are gradually relaxed, it would indeed be equivalent to a "substantial tax reduction" because companies would no longer need to spend large amounts of money to meet import thresholds or be forced to exit the market. Wood further pointed out that the core of this negotiation is:
"To reduce unreasonable trade barriers, which can be considered an indirect tax reduction and market liberalization."
Can tax cuts save the fiscal deficit? Wood says that combined with economic growth, it may not necessarily worsen the deficit
Facing the continuously expanding US fiscal deficit, Wood stated:
- Previously, Trump reduced corporate tax from 35% to 21%, causing considerable controversy
- However, total corporate tax revenue increased from $200 billion to $500 billion, showing that tax cuts can stimulate economic recovery
- After the tax cut, companies are willing to set up factories in the US, increase investment, and no longer seek ways to avoid taxes or transfer money overseas.
She also optimistically views the production growth driven by AI, believing that future GDP growth might be far higher than expected:
"Not by cutting expenses, but by using technological innovation to fill the deficit gap."
Can US innovation continue? The key lies in technology and government efficiency
Wood pointed out that previously, the Biden administration's conservative approach to digital assets almost drove innovation away. However, now, with the emergence of new technologies like AI, many innovative talents and businesses are returning to the US.
She emphasized that currently:
"AI training costs decrease by 75% annually, and inference costs even drop by 85% to 95%, which can significantly lower the innovation threshold and stimulate AI development."
She also noted:
- Palantir, an AI data analysis software provider: Assists the US Department of Defense with advanced data integration and analysis tools.
- US Food and Drug Administration (FDA): Adopts AI systems to diagnose and eliminate traditional animal experiments.
Demonstrating these as practical examples of "technology improving government efficiency".
Biotechnology, digital assets, and data science innovation stocks are about to take off
Wood also stated in the interview that the following innovation stocks are undervalued, such as:
CRISPR Therapeutics: Using gene editing technology to treat sickle cell anemia
Palantir: Currently actively assisting US defense.
Coinbase: Benefiting from the recovery of the digital asset market.
Roku: Becoming the main smart TV operating system.
Twist Therapeutics: Developing blood testing tools that may detect cancer early.
She emphasized that ARK investments have faced valuation pressure in recent years, and now the valuation has returned to near market average, making the oversold innovation stocks a buying opportunity.
Biotech innovation is the biggest beneficiary of AI, traditional pharmaceutical industries should be cautious of being eliminated
Wood believes that the biotechnology industry will be the biggest beneficiary of AI. She stated that AI can help:
Reduce drug development time from 13 years to 8 years.
Detect disease risk in advance, shifting from post-treatment to pre-prevention.
Replace long-term medication with one-time therapy, such as CRISPR gene editing technology, which may replace commonly used cholesterol-lowering drugs (Statins).
However, she also warned that traditional pharmaceutical companies that do not embrace AI and gene technology may easily be replaced by new biotech companies.
Be cautious of AI investment bubble, biotech and medical industries are the biggest beneficiaries of AI
At the end of the interview, Wood also reminded everyone to be careful of the AI investment boom. She stated:
"When a large amount of capital flows into a specific industry, it may simultaneously increase the risk of bubble formation."
She believes that the current market is too focused on AI infrastructure, such as data centers and chip manufacturers, while overlooking the biotech and medical industries that will truly generate revolutionary changes through AI applications.
Risk Warning
Cryptocurrency investment carries high risk, and its price may fluctuate dramatically. You may lose all your principal. Please carefully assess the risks.