The Public Markets AI Boom Isn’t Like The Dot-Com Era – Issue #17
Spotlight: Public Markets AI Boom Isn’t Like The Dot-Com Era
Every time a tech cycle hits a new gear, we start hearing the “Dot-Com” comparisons. It’s an easy headline, but it misses the reality on the ground.
Sure, AI gains are moving fast. Enthusiasm is through the roof, and capital spending is massive. But here’s the difference: today’s AI leaders are building on actual revenue, not vibes and vision decks.
Goldman Sachs research on Public markets shows just how different this moment is from 25 years ago:
Valuations are well below prior bubble extremes.
Margins are meaningfully higher.
Most investment is being funded from operating cash flow, not leverage.
Capex is rising but remains well below historical infrastructure boom levels.
Monetization isn’t a “someday” goal; it’s happening in the near-term.
This isn’t a narrative chase. It’s an earnings-backed platform transition.
But for founders, a favorable market doesn’t mean a “soft” one. We’ve moved past the era of “multiple expansion,” where simply being in the right sector was enough. Today, the market is structurally disciplined; it rewards velocity of learning, defensible workflow moats over just being wrappers on models, and products that pass the “Pain Test”—becoming indispensable rather than just a flashy add-on. In this cycle, the market demands the unit economics and mission-critical utility that prove a business is durable.
The Mayfield View: We’re backing AI founders building businesses that prioritize durable workflows and real-world adoption over “growth at any cost.” In a world of ubiquitous models, enduring value comes from fundamentals and human-centric design, not just a good story.
The Bottom Line for Founders: This is a strong but disciplined market. The days of “easy” gains from sector tailwinds are over. It’s no longer about the narrative; it’s about the execution.
Signals Shaping the Future of AI:
Infrastructure
Anthropic pledges to cover grid upgrade costs for AI data centers. Anthropic said it will pay 100% of required grid infrastructure upgrades and invest in new power sources to offset electricity demand from its AI data centers as U.S. capacity tightens. Click here
Mistral commits €1.2 billion to Swedish AI data center expansion. Mistral announced plans to invest €1.2 billion in building out AI data center capacity in Sweden to support model training and inference workloads. Click here
Inference providers report up to 10x cost reductions on NVIDIA Blackwell. Baseten, DeepInfra, Fireworks AI, and Together AI reported 4x to 10x reductions in cost per token by running open-source models on NVIDIA’s Blackwell platform with optimized inference software stacks. Click here
OpenAI deploys Cerebras chips for low-latency codings model. OpenAI launched GPT-5.3-Codex-Spark, optimized for near-instant code generation and running on Cerebras hardware, marking its first major inference partnership beyond NVIDIA GPUs. Click here
Cisco launches Silicon One G300 AI networking chip. Cisco unveiled a new 3nm switch chip designed to accelerate data movement across large AI data centers, competing with Broadcom and NVIDIA in AI infrastructure networking. Click here
Samsung begins mass production of HBM4 for AI data centers. Samsung announced commercial shipments of its HBM4 memory, delivering up to 11.7 Gbps speeds, 3.3 TB/s bandwidth per stack, and improved power efficiency and thermal performance for next-generation AI infrastructure. Click here
Enterprise
Amazon steers teams toward in-house ‘Kiro’ as employees push for Claude Code. Business Insider reports that Amazon is directing engineers to use its internal AI coding assistant, Kiro, for production work, prompting internal lobbying from roughly 1,500 employees who favor Anthropic’s Claude Code. Click here
Spotify says top developers now rely entirely on AI coding tools. Spotify executives said its highest-performing engineers have not written code manually since December, instead using internal AI systems powered by Claude Code to generate, test, and deploy features. Click here
OpenAI begins testing ads in ChatGPT’s Free and Go tiers. OpenAI announced it is rolling out advertising to U.S. users on its Free and $8 per month Go plans, while paid tiers, including Plus, Pro, Business, Enterprise, and Education, will remain ad-free. Click here
Two more xAI co-founders depart, bringing total exits to six. Yuhuai (Tony) Wu and Jimmy Ba announced their departures from xAI, meaning half of the company’s 12-person founding team has now left since launch. Click here
Alphabet expands bond sale to more than $30 billion. Alphabet increased its global debt offering to over $30 billion amid strong investor demand, as the company prepares for up to $185 billion in capital expenditures tied to its AI infrastructure expansion. Click here
Microsoft advances in-house AI development while maintaining OpenAI ties. The company is building its own foundation models and custom AI chips to reduce reliance on OpenAI while preserving its long-term partnership and Azure exclusivity agreement through 2032. Click here
Capital Flows
Apptronik raises $520 million to scale humanoid robot production. Apptronik secured $520 million in new funding to expand manufacturing and development of its AI-powered humanoid robots. Click here
Simile raises $100 million to build predictive AI for human behavior. Simile secured $100 million in funding, led by Index Ventures, to develop models that anticipate human actions, including forecasting likely questions during corporate earnings calls. Click here
Runway raises $315 million at a $5.3 billion valuation. AI video startup Runway closed a $315 million Series E round to expand its world model research, video-generation platform, and compute capacity. Click here
Blackstone increases Anthropic stake to approximately $1 billion. Blackstone invested an additional $200 million in Anthropic, bringing its total stake to about $1 billion at a valuation of roughly $350 billion as part of the company’s ongoing funding round. Click here
Former GitHub CEO raises $60 million seed round for AI code management startup. Thomas Dohmke’s new startup, Entire, secured $60 million at a $300 million valuation to build open-source tools that help developers manage and review software generated by AI coding agents. Click here
Anthropic closes $30 billion funding round at $380 billion valuation. Anthropic announced it raised $30 billion at a $380 billion post-money valuation, with participation from Coatue, GIC, Microsoft, and NVIDIA, to fund model training, infrastructure expansion, and enterprise product development. Click here
Research
Microsoft AI CEO predicts widespread automation of white-collar tasks. Microsoft AI chief Mustafa Suleyman said he expects AI to reach human-level performance across most professional computer-based tasks within 12 to 18 months. Click here
MiniMax releases M2.5 coding model with 80.2% SWE-Bench score and lower pricing. MiniMax introduced M2.5 and M2.5-Lightning, reporting 80.2% on SWE-Bench Verified, improved multi-step tool use, and pricing starting at $0.30 per million input tokens and $2.40 per million output tokens. Click here
Alibaba releases open-source robotics foundation model RynnBrain. Alibaba’s DAMO Academy introduced RynnBrain, an open-source model designed to help robots perform real-world tasks by understanding spatial environments and sequencing actions. Click here
Mathematicians publish research-level AI benchmark paper. Researchers, including Martin Hairer, Mohammed Abouzaid, and Lauren Williams, released “First Proof,” a paper introducing unpublished research math problems to evaluate how large language models perform on genuine research-level questions. Click here
Policy
OpenAI accuses DeepSeek of distilling US AI models. OpenAI told U.S. lawmakers that Chinese AI firm DeepSeek used distillation techniques and evasive methods to extract outputs from leading U.S. models to train its R1 chatbot. Click here
White House accelerates AI adoption across federal agencies. The Trump administration directed federal departments to rapidly deploy AI across policing, health care, defense, science, and other operations, removing internal restrictions to expand use of the technology. Click here
India requires social media platforms to remove unlawful content within three hours. India introduced new rules mandating that major platforms remove illegal content within three hours of notification, down from 36 hours, and requiring labeling and traceability measures for AI-generated content. Click here
MPA criticizes ByteDance’s Seedance 2.0 over copyright concerns. After AI-generated videos featuring likenesses of Tom Cruise and Brad Pitt went viral, the Motion Picture Association accused ByteDance’s new Seedance 2.0 video model of widespread unauthorized use of copyrighted works. Click here
The Trump administration pauses planned restrictions on Chinese tech. Reuters reports that the White House has shelved proposed bans on Chinese telecom and data center equipment ahead of an April summit with President Xi Jinping. Click here
Global AI Strategy
European chip leaders highlight global reliance on EU technology. In POLITICO, executives from ASML and Imec, alongside EU officials, said the U.S. and other regions depend on European chipmaking equipment as the bloc advances its semiconductor strategy and prepares a second Chips Act. Click here
Microsoft expands AI footprint in Australia with $5 billion data center investment. Satya Nadella’s upcoming visit to Sydney coincides with a $5 billion AI data center expansion, part of Microsoft’s broader push to grow AI infrastructure capacity across Asia-Pacific markets. Click here
Google warns EU sovereignty push could undermine competition. In the Financial Times, Google’s Kent Walker said proposed EU tech sovereignty measures risk limiting access to foreign technology as Brussels prepares its digital sovereignty package. Click here
Talent Signals
Each week, we spotlight key roles tied to the themes shaping this week’s AI headlines, connecting talent to the companies driving the news.
@Runway builds generative video and multimodal AI systems used by creators, studios, and enterprises to produce high-quality content at scale. As AI tools move from experimentation into real production workflows across media and design, Runway is hiring across research, engineering, and product teams. Open roles are listed on its careers page. Click here
@Apptronik develops AI-powered humanoid robots designed for real-world tasks in manufacturing, logistics, and other industries. As the robotics sector attracts major investment and scales toward commercialization, Apptronik is hiring across technical and product teams; open roles are on its careers page. Click here
@InworldAI builds real-time multimodal AI that powers dynamic characters and interactive experiences across games, AR/VR, and virtual worlds. The company is hiring across engineering, product, and platform roles as developers increasingly embed AI interaction layers into apps. Open roles are listed on its careers page. Click here
You can see all the opportunities at Mayfield-backed AI companies here, and across the broader ecosystem here.
Social Signals
The most important conversations in AI are unfolding across social media, where top voices are shaping the next wave of signals and strategy. Here are some of the top social signals and their takes from the past week.
Matt Shumer (Click here) — “Something Big Is Happening. I think we’re in the ‘this seems overblown’ phase of something much, much bigger than Covid. On February 5th, two major AI labs released new models on the same day, GPT-5.3 Codex and Opus 4.6. And something clicked. I am no longer needed for the actual technical work of my job.” In a post that has now surpassed 80 million views, Shumer argues that recent model releases mark a step change, not an incremental upgrade, describing AI systems that can independently write, test, debug, and iterate on complex software. He frames the acceleration as already visible inside tech, with AI contributing to building the next generation of itself and rapidly expanding the length and complexity of tasks it can complete. The viral reaction underscores how sharply public attention is shifting toward the pace, scale, and economic implications of frontier AI.
Lenny Rachitsky (Click here) — “AI is writing virtually all code at OpenAI. 95% of the engineers use Codex… engineers who embrace these tools open 70% more pull requests than their peers… The role of a software engineer is shifting from writing code to managing fleets of AI agents.” Rachitsky shares takeaways from Sherwin Wu, describing how OpenAI has moved to an agent-first workflow, with Codex reviewing every PR before humans and compressing review times from 10 to 15 minutes to 2 to 3. He notes that top performers are becoming disproportionately more productive, enterprise ROI depends on bottom-up adoption, and startups should build for where models are going, not where they are today. The thread frames AI not as incremental tooling, but as a structural shift in how software is built, who captures the upside, and how value accrues in an AI-augmented economy.
John Carmack (Click here) — “The modern age has richly rewarded people with a combination of high intelligence and high agency. Now that many aspects of intelligence are successfully being automated, it seems likely that people with relatively lower intelligence but exceptional agency will come into their own if they are willing to egolessly accept AI advice. Imagine a ruthless criminal who completely trusts everything their always-on AI glasses are telling them.” Carmack argues that as intelligence becomes automated, relative advantage may shift toward individuals who act decisively and defer to AI systems without ego. He sketches a future where AI functions as an always-on strategic layer, amplifying agency regardless of baseline intelligence, and raising questions about how such augmentation changes power dynamics across society.
