The AI Gold Rush: Money Today Is in Infrastructure – Issue #27
Spotlight: The Future of Sales and Role of the CRO
In every gold rush, the miners get the headlines. But the enduring fortunes are built by the people providing the picks and shovels.
We are living through the AI gold rush. The first wave of massive value creation has already accrued to the infrastructure layer: NVIDIA on compute, Micron on memory, Microsoft/AWS/Google on cloud, OpenAI and Anthropic on foundation models.
A second infrastructure wave is forming underneath the first. As AI scales to tens of thousands of GPUs, the bottlenecks are shifting from chips to physics:
– Power infrastructure
– Advanced liquid cooling systems
– Optical and next-gen interconnects
– Disaggregated and pooled memory architectures
– Scale-up switching architectures
The AI gold rush isn’t just digital. It’s physical.
Three structural forces explain why value first concentrates at the infrastructure layer.
Capital intensity creates oligopolies.
Scale compounds. Bigger clusters produce better models, which generate more revenue, which fund even larger clusters
Ecosystem lock-in. Once infrastructure becomes default infrastructure, it compounds.
The consolidation of the AI infrastructure layer isn’t a threat – it’s a signal. Infrastructure consolidates first, then innovation explodes, and founders win by solving bottlenecks and owning workflows and customer relationships.
This week made that shift explicit. Meta signed a multibillion-dollar deal to use Amazon’s Graviton chips for AI workloads, OpenAI committed over $20 billion to Cerebras to secure non-NVIDIA compute supply, Google moved to co-develop next-generation AI chips with Marvell, and NVIDIA-backed VAST Data raised $1 billion at a $30 billion valuation to scale AI data infrastructure.
Full Weekend Edition below. 👇
Signals Shaping the Future of AI:
Infrastructure
Meta signs multibillion-dollar deal to use Amazon’s Graviton chips for AI workloads. The agreement gives Meta access to large-scale CPU capacity for inference and supporting tasks alongside GPUs in its AI infrastructure. Click here
OpenAI agrees to spend more than $20 billion on Cerebras chips over three years, in a deal that could also give it an equity stake. The commitment secures a large non-NVIDIA supply of AI compute and accelerates diversification of the high-end training hardware ecosystem. Click here
Google is in talks with Marvell to co-develop two new AI chips, including a memory processing unit to pair with TPUs and a new inference-focused TPU. The partnership targets more efficient, lower-cost serving of large models and extends Google’s in-house silicon roadmap beyond existing TPU generations. Click Here
Nokia raises its AI business growth targets after a 54% jump in quarterly operating profit, driven by demand for optical gear in AI data centers. The performance shows how hyperscaler build-outs are reshaping telecom equipment markets and shifting value toward high-capacity transport. Click Here
Enterprise
OpenAI launches GPT-5.5, its newest model for coding, research, and computer-based work. The release expands agentic capabilities across software engineering, document creation, data analysis, and tool use, with rollout starting in ChatGPT and Codex. Click here
Microsoft plans to embed Anthropic’s Claude Mythos Preview into its Security Development Lifecycle to scan code for vulnerabilities and speed remediation. Using a frontier-grade cybersecurity model in core engineering workflows signals how large software vendors are operationalizing high-risk AI under controlled programs like Anthropic’s Project Glasswing. Click Here
Google rebrands its cloud AI stack under “Gemini Enterprise” and makes AI agents the centerpiece of its enterprise strategy at its annual cloud conference. The company is positioning agentic systems, backed by new TPU 8t and 8i chips and expanded governance features, as production infrastructure for large business customers. Click Here
Adobe launches CX Enterprise, an AI-agent suite for corporate digital marketing that automates and personalizes customer interactions. The product, integrated with Amazon, Microsoft, Anthropic, OpenAI, and NVIDIA platforms, shows incumbents using agentic AI to defend and extend high-margin marketing software franchises. Click here
Freshfields partners with Anthropic to deploy Claude across global legal workflows. The firm is rolling out AI tools to 5,700 employees and co-developing agentic systems for contract review, research, and multi-step legal tasks. Click here
Capital Flows
Google plans to invest up to $40 billion in Anthropic to expand AI capacity. The deal includes an initial $10 billion commitment with additional funding tied to performance milestones and infrastructure growth. Click here
NVIDIA-backed VAST Data raises about $1 billion at a $30 billion valuation to scale its AI data platform. The round, led by Drive Capital and Access Industries, underscores investor belief that data and storage software providers will capture outsized value from AI infrastructure growth. Click Here
Jeff Bezos–backed Project Prometheus is close to raising $10 billion at a $38 billion valuation to build AI for engineering and manufacturing. The funding, with JPMorgan and BlackRock among investors, adds another heavily capitalized frontier-AI competitor focused on industrial workflows. Click Here
OpenAI is set to invest up to $1.5 billion into DeployCo, a $10 billion joint venture with private equity firms that will guarantee investors a 17.5% annual return. The vehicle is designed to finance large-scale enterprise deployments of OpenAI’s workplace tools, blending AI vendor economics with private equity distribution. Click Here
Sooth Labs, founded by ex-Meta researchers, raises about $50 million at a valuation of approximately $335 million to build AI models that forecast geopolitical and market events. Backing from Yann LeCun, Jeff Dean, and others highlights investor appetite for specialized, high-value forecasting models for institutional clients. Click Here
Research
Sony AI’s Ace robot becomes the first autonomous system to achieve expert-level play against elite and professional human table tennis players, with results published on the cover of Nature. The system combines advanced sensing, reinforcement learning, and high-speed control to operate at human reaction times in a complex physical sport, marking a major step for embodied AI. Click here
OpenAI releases Privacy Filter, an open-weight, on-device model for detecting and redacting personally identifiable information in text under an Apache 2.0 license. The 1.5-billion-parameter model supports 128k-token contexts and provides a research-grade, privacy-focused component that can be embedded into training, logging, and review pipelines across AI systems. Click Here
Tencent unveils its Hy3 foundation model as its most powerful AI system to date and rolls it out across chatbots, coding tools, and consumer apps. The upgrade is Tencent’s first major test of its revamped AI leadership and is a key move for China’s largest internet firm to compete with US labs and domestic rivals. Click Here
Policy
Anthropic CEO Dario Amodei meets the White House chief of staff amid a Pentagon dispute over the use of the Mythos model, according to Axios. The engagement highlights tensions between US defense agencies and a leading lab over access to high-risk frontier models and shows Mythos is now a subject of direct executive-branch attention. Click Here
Germany’s central bank chief calls for broad access to Anthropic’s Mythos model. Bundesbank President Joachim Nagel said financial institutions and regulators should be able to test the model as concerns grow over its ability to identify and exploit software vulnerabilities. Click Here
Clarifai deletes 3 million OKCupid user photos and facial-recognition models trained on them after an FTC privacy settlement with the dating site. Forcing destruction of both source data and trained models shows US regulators are willing to reach into AI pipelines, not just upstream data practices, when enforcing privacy rules. Click Here
Global AI Strategy
The White House accuses China of “industrial-scale” theft of US AI intellectual property via coordinated distillation campaigns targeting frontier models. The memo raises the stakes in the US–China tech conflict, links model-stealing to national security, and could affect future export approvals for NVIDIA AI chips to China. Click Here
Taiwan launches a financial-sector large language model project involving 16 institutions under the Taiwan FinTech Alliance. By building domain-specific models tuned to local regulations and practices, Taiwan is seeking to reduce dependence on foreign AI platforms and strengthen its financial and technological sovereignty. Click Here
Volkswagen Group unveils a China AI roadmap that will equip new locally built vehicles with in-car AI “agents” starting in the second half of 2026. The strategy is aimed at catching up with Chinese EV competitors on intelligent cockpit and assistant capabilities in the world’s largest auto market. Click Here
OpenAI briefs US federal agencies, state governments, and Five Eyes countries on its GPT-5.4-Cyber defensive security model. The outreach positions frontier AI models as national security tools and integrates labs more tightly into government cyber defense planning and procurement. 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.
@Sierra the AI agent company founded by Bret Taylor and Clay Bavor, is deploying conversational agents across Fortune 500 customer-facing teams. Sierra closed a $350 million round at a $10 billion valuation and is hiring aggressively across forward-deployed engineering, research, and product. Open roles on its careers page. Click here
@Cognition, the team behind Devin, is in talks to raise at a $25 billion valuation as its AI software engineer scales inside enterprise engineering orgs. The company is hiring across research, forward-deployed engineering, and product. Open roles on its careers page. Click here
@NeuBird closed a $19.3 million round in April, scaling its agentic AI platform for site reliability and enterprise production operations. Open roles 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.
Aaron Levie (Click here) — “AI makes it easy to explore more than you did before, and so you start doing far more as a result. I regularly have small things that end up consuming hours because the agent made it easy to get started, but you still have to do the rest of the work to complete the project.” Levie argues that AI is expanding the surface area of work rather than reducing it, lowering the cost to start tasks that previously would not have been attempted. As a result, individuals and teams take on more projects, turning latent ideas into active work and increasing overall output rather than decreasing effort.
Alex Lieberman (Click here) — “Historically, leaders had to think about allocating budget for headcount and software. Now, the token budget must be a serious consideration. The models are good enough now that most bad output is user error, not technology error. A skill is nothing more than an SOP, a repeatable process that can be turned into a prompt.” Lieberman shares takeaways from training senior executives, highlighting how AI adoption is shifting from tools to operating models. The focus is moving toward budgeting usage, managing context, codifying workflows, and redesigning processes, with agents and systems like Claude Cowork making these capabilities accessible beyond technical teams.
