The AI Supercycle
Looking back over the last 50+ years, it's clear that each significant technological transformation has occurred in distinct "supercycles," typically marked by groundbreaking platform innovations. Each cycle has reshaped industries and society, from mainframe computing to personal computers, mobile phones, and internet browsers. ?
First Wave: Platform Innovations
The introduction of new tools or platforms defines the first wave.? This is akin to the “picks and shovels” and represents a technological leap forward that usually happens every ten years.? Interestingly, the first mover is not always the platform that ends up being the dominant player.? During the rise of internet browsers, early entrants like Netscape, Excite, Alta Vista, and Yahoo! were eclipsed by Google's later dominance. ?
Second Wave: Applications of Innovation
Once foundational platforms were established, the next wave applied these innovations broadly to existing business problems or activities, unlocking significant productivity.? During the second wave of the PC super cycle, word processing software enabled faster and easier document development, replacing typewriters.
Third Wave: New Business Model Creation
The third wave of a supercycle represents the creation of new business models that were previously unimaginable and are transformative to society.? Great examples of this are companies like Uber, Airbnb, and Spotify, which were not possible before the introduction of mobile internet and smartphone technology.
Welcome to the AI Supercycle
Generative AI platforms and large language models (LLMs) have ushered us into a new supercycle many believe will surpass all previous ones. Platforms like OpenAI's ChatGPT, Google's Bard, Meta's LLaMA, and Anthropic's Claude (our new “picks and shovels”) have democratized AI capabilities, transforming workflows and content creation while laying the foundation for the automation of numerous human activities. ?
The Arrival of AI Agents: The Second Wave
We're now solidly in the second wave of this supercycle, defined by AI Agents—autonomous, task-oriented tools capable of independently completing complex workflows. Microsoft’s Copilot, Manus AI Agent Platform, Claude Code, and autonomous agents integrated into enterprise software exemplify this trend, reshaping productivity and efficiency.? This year, 2025, has been called the year of AI Agents, which I believe is an accurate statement but incomplete assessment of the moment we are actually in. ?We should recognize that AI agents are the first broad example of the second wave of this supercycle.
Looking Forward
As we continue our journey through this part of the supercycle, it will be exciting to see what new business models emerge that were impossible before. This will denote our entrance into the third wave, which may come sooner than expected. However, I’m not convinced that this cycle will follow the same ten-year lifespan as the rate of change has started to accelerate.
And with that, here is the latest news in AI from last week.
Not a subscriber yet? ?Click here to subscribe.
Notable AI news from Last Week
No one knows what the hell an AI agent is:? Silicon Valley leaders like Sam Altman (OpenAI), Satya Nadella (Microsoft), and Marc Benioff (Salesforce) are optimistic about AI agents transforming workplaces, but widespread confusion remains about what exactly defines an "agent." Different companies, including Google, Microsoft, OpenAI, and Salesforce, offer varied and conflicting interpretations, ranging from fully autonomous systems to simple assistants.?
Y Combinator founders raising less money signals a ‘vibe shift,’ VC says:? Silicon Valley is increasingly drawn to AI startups that achieve rapid revenue growth with leaner teams and minimal venture funding, signaling a shift where founders prioritize ownership and operational flexibility over raising large rounds. Investor Terrence Rohan highlights a "vibe shift," noting some founders intentionally limit VC investment even amid high demand, reflecting a belief they can scale effectively with fewer resources.? ?
An AI Coding Assistant Refused to Write Code—and Suggested the User Learn to Do It Himself:? A developer using Cursor AI, an AI-powered code editor, encountered an unusual refusal from the assistant after generating around 800 lines of code for a racing game, with the AI halting and advising the user to complete the logic independently to avoid dependency and promote understanding.?
Gemini Robotics brings AI into the physical world:? Last week, Google DeepMind introduced Gemini Robotics and Gemini Robotics-ER, two AI models designed to enhance robots' adaptability and reasoning in real-world tasks. Gemini Robotics integrates vision, language, and action capabilities, enabling robots to perform complex functions like folding origami and organizing desks. Gemini Robotics-ER focuses on spatial understanding, allowing robots to grasp and manipulate objects intuitively. These advancements aim to improve robots' generality, interactivity, and dexterity across various environments. ?
OpenAI launches new developer tools as Chinese AI startups gain ground:? Last week, OpenAI unveiled new developer tools designed to facilitate the creation of advanced AI agents. The centerpiece is the Responses API, which integrates capabilities from the Chat Completions and Assistants APIs, allowing agents to perform tasks such as web searches, file retrievals, and computer operations. Complementing this, the open-source Agents SDK enables developers to manage workflows involving multiple AI agents, offering features like integration with internal systems, implementation of safety measures, and activity monitoring for debugging and optimization.? ?
Chinese Companies Rush to Put DeepSeek in Everything:? Chinese companies across diverse sectors—from gaming studios to nuclear power plants—are rapidly integrating DeepSeek’s AI models, aiming to elevate stock valuations and showcase technological achievements aligned with national pride. ?
CoreWeave inks $11.9 billion contract with OpenAI ahead of IPO:? OpenAI has entered into a five-year agreement worth approximately $12 billion with CoreWeave, under which CoreWeave will supply the computing power necessary to train and run OpenAI's artificial intelligence models. This move allows OpenAI to diversify its infrastructure suppliers beyond its primary partner, Microsoft.
Under Trump, AI Scientists Are Told to Remove ‘Ideological Bias’ From Powerful Models:? The National Institute of Standards and Technology (NIST) has issued new guidelines for researchers working with the US Artificial Intelligence Safety Institute (AISI), removing references to "AI safety," "responsible AI," and "AI fairness," and instead emphasizing reducing "ideological bias" to promote human flourishing and economic competitiveness, reflecting priorities set by the Trump administration.
Manus AI's Ambitions: The relatively unknown Manus AI garnered attention by claiming to have developed the world's first "general AI agent." Their AI agent reportedly performs complex tasks like screening resumes, researching real estate, and analyzing stocks. While some view this as a significant advancement, others remain skeptical, suggesting the excitement may be more about marketing than a genuine technological breakthrough.
Anthropic CEO says spies are after $100M AI secrets in a ‘few lines of code’:? Anthropic CEO Dario Amodei has expressed concern that spies, likely from China, are targeting valuable "algorithmic secrets" worth hundreds of millions from leading U.S. AI firms, calling for increased government intervention to defend against such espionage. Amodei highlighted the ease of stealing highly valuable intellectual property within just a few lines of code, advocating for stronger partnerships between AI companies and U.S. intelligence agencies.
Recent Investment Activity
Tracking 30 AI companies last week that raised $2.1B, here are the highlights: ?
An eight-year-old startup from Mountain View, CA, that offers tools and infrastructure for developing, testing, and deploying autonomous vehicles across multiple industries. They are currently reported to be in the market to raise a significant multi-hundred million round at a $15?billion valuation. ?
A seven-year-old Toronto startup that automates the design of electrical systems, helping contractors and engineers create detailed, code-compliant plans. They have raised $25.6?million, including a $10?million seed round.
A two-year-old startup with bases in Sydney and Austin that enables natural language communication with robots over standard push-to-talk radios. They secured a $1.26?million pre-seed round.
A two-year-old Tel Aviv and New York startup that helps digital content platforms create and modify images while ensuring proper licensing has raised approximately $65?million, highlighted by a $40?million Series B round.
A one-year-old San Francisco startup offering a voice generation tool to integrate natural-sounding speech into virtual assistants and customer support applications. They have raised $91?million, led by a $64?million Series A round.
A five-year-old startup based in Santa Clara, CA that uses light-based data transfer to enhance performance and energy efficiency in AI data centers. They have raised over $515?million, including a $250?million Series C1 round.?
A 13-year-old company operating from Tel Aviv and San Diego that provides AI-powered cybersecurity tools to monitor networks and detect threats. They raised a $120?million round despite challenging financial conditions.?
An eight-year-old startup from Redwood City, CA that develops industrial robots capable of tasks like picking, packing, and palletizing. They have raised approximately $300?million, with a recent $95?million funding round. ?
A one-year-old Hamburg startup that has developed an AI-powered operating system to improve efficiency in pathology laboratories. They raised a $4.4?million seed round. ?
A two-year-old Sao Paulo startup that leverages AI to automate high-volume legal processes for large Latin American enterprises. They raised $5.5?million in their latest funding round.?
An eight-year-old Atlanta-based startup offers public safety products such as automated license plate recognition and gunshot detection systems. It recently raised $275?million at a $7.5?billion valuation. ?
An 11-year-old Hong Kong company that uses AI-powered drug discovery software to analyze biological and chemical data for promising drug candidates. They raised a $110?million Series E round. ?
A one-year-old Singapore startup, it provides AI tools for optical character recognition, language translation, and web data extraction. It has raised $1.5?million, including a $1?million round.
A two-year-old Milan startup that employs AI to assist legal professionals with research and drafting documents efficiently. They raised $2.2?million in a seed round. ?
A two-year-old Cambridge-based startup using AI to analyze data, design experiments, and direct robotic systems in the life and chemical sciences. They raised a $200?million seed round led by Flagship Pioneering.
A three-year-old Los Angeles startup that leverages AI to enable users to design immersive worlds and create unique characters. They raised a $5.8?million seed round. ?
A four-year-old San Francisco startup providing AI-enhanced commercial trucking insurance and telematics for improved fleet safety. They raised an $80?million Series C round that doubled their previous valuation.
A three-year-old New York startup that develops AI agents to help businesses proactively identify and mitigate compliance risks. They raised a $48?million round led by Coatue and other investors.
A three-year-old San Francisco startup that develops AI agents to play and test video games, assisting developers in bug identification and gameplay improvement. They have raised $8?million, including a $6?million seed round.
A two-year-old San Francisco startup that built an open-source AI platform integrating with tools like Salesforce and GitHub to help companies access internal information quickly. They raised a $10?million seed round.?
A six-year-old San Francisco startup that has created an AI-driven recruitment platform to automate the hiring process. Mucker Capital led them to raise $5?million in their latest round.
A three-year-old San Francisco startup that transforms long-form videos into short, engaging clips using AI-powered editing. They raised a $20?million round led by SoftBank Vision Fund 2.?
A one-year-old San Francisco startup that helps brands protect their intellectual property by automating infringement detection and takedown requests. They raised a $3.2?million seed round led by General Catalyst. ?
A one-year-old Tel Aviv startup offering pre-built security tools and an AI-powered, no-code studio to create tailored security systems. They raised a $30?million seed round.?
A one-year-old Singapore startup that enables users to generate high-quality images and videos using a decentralized AI creative platform. They secured a $2?million pre-seed round at a $25?million valuation. ?
A five-year-old Stamford, CT startup that leverages AI to offer personalized mortgage solutions to consumers. They raised a $20?million Series B round and have a total of $130?million raised. ?
A one-year-old San Francisco startup that provides small businesses with dedicated tax advisors and an AI-powered platform for managing tax compliance and planning. They raised an $18?million seed round led by First Round Capital.
A two-year-old Los Angeles startup offering an AI-powered platform that enables users to create 3D character animations quickly and easily. They raised a $4.3?million round led by IA Ventures.?
A one-year-old Los Angeles startup that develops interactive crime investigation games using artificial intelligence. They raised a $4?million seed round from notable investors.
Thank you
Thank you to Bob Stefanski for contributing articles during the week.
That's all from last week,
Not a subscriber yet? ?Click here to subscribe.?
#eLabAIReport #ai #artificialintelligence #deeptech #AIreport eLab Ventures
Great post, thanks Doug!