In partnership with

The Commodization of LLMs

If 2023 was the year of the LLM arms race, 2025 is the year of strategic clarity. As generative AI matures, the field is seeing a new reality: most serious builders are leveraging the same foundational models such as OpenAI, Anthropic, Google, Meta & Mistral. According to ICONIQ’s 2025 State of AI report, the average AI company now juggles nearly three models in production, selecting between them for cost, latency, or geographic fit.

This rapid standardization has fueled innovation but also exposed a hard truth: the model layer is no longer a defensible moat. Fine-tuning, open weights, and seamless model-swapping are now table stakes. Last year’s differentiators, exclusive access to GPT-4, Claude, or Gemini are today’s baseline.

So, where will tomorrow’s winners carve out their edge?

1. Composable Architectures & Multi-Model Orchestration

Leading teams are architecting for flexibility, not vendor lock-in. They’re designing systems that can hot-swap models based on task, price, or performance. This resilience shields them from pricing shocks, API downtime, or regulatory shifts, and lets them optimize ROI for each use case.

For example, rather than relying on a single foundational model, many teams now incorporate platforms like Perplexity as one node among several in their multi-model stacks. It uses real-time web search integration and support for multiple advanced models.

This shift is fueling a new ecosystem: orchestration layers, agent frameworks, and vector-native databases are becoming critical infrastructure. The competitive advantage isn’t about having the “best” model it’s about routing, monitoring, and optimizing across models faster than anyone else.

2. Verticalization & Domain Expertise

As horizontal model performance converges, defensibility is shifting to vertical focus. High-growth AI startups are far more likely to build or fine-tune models for specific domains legal, healthcare, security, finance where regulatory alignment and specialized data matter most.

Domain expertise, compliance, and proprietary training data are now the real levers. Investors should zero in on teams solving complex, regulated problems at the intersection of AI and industry.

3. Proprietary Data & Feedback Loops

Data remains one of the last sustainable moats. Teams that own unique, high-quality datasets or generate them through product usage are best positioned to train differentiated models or drive outcomes with reinforcement learning.

But it’s not just about data ownership. The fastest-scaling companies are those with tight feedback loops: rapid deployment, real-world measurement, and continuous improvement. Closed feedback cycles and performance telemetry at every layer are the new gold standard.

4. Inference Efficiency & Margin Discipline

API fees and inference costs are now the most stubborn line items in the AI P&L. This has sparked a wave of experimentation with model distillation, quantization, and open-source alternatives like DeepSeek and Mistral.

The most disciplined teams are starting to look less like flashy AI startups and more like performance-obsessed devtools companies. As investors scrutinize margins and unit economics, startups that deliver robust performance while slashing inference spend will stand out.

5. Explainability, Trust, and User Experience

Even the best technical output must be usable and trustworthy. Trust, transparency, and control are becoming core product features. Vertical AI teams are embedding explainability and decision traceability into their UIs especially in compliance-heavy sectors turning regulatory hurdles into UX differentiators.

With some like Unite.AI asking the question: Are AI Models becoming Commodities?

The Bottom Line

In a market where everyone has access to the same foundation models, the advantage won’t go to those who buy the best, it’ll go to those who build the smartest stack around it. Investors should look beyond polished demos and ask sharper questions: Can this team switch models without friction? Are inference costs under control? Is their data truly defensible? How fast do their feedback loops run? The winners won’t just deploy AI, they’ll operationalize with velocity, infrastructure, and iteration built to outpace the competition.

Bay Area Startups Collectively Secured $7.4B in JULY MTD

$7.4B went to Bay Area startups so far in July, with the bulk of that going into AI, primarily frontier model and AI infrastructure companies, including xAI and Perplexity AI, and SuperAnnotate AI this week.

High tech manufacturing on steroids: California Forever announced that their planned community will include the Solano Foundry, a 2,100-acre, 40 million-square-foot area that will be the nation's largest advanced manufacturing park that will provide 90-day permitting; affordable power generation, over 2 GW of renewable and baseload power and over 5,000 GWh of battery storage; clustered industries, so that R&D, prototyping, and scaled production can be on a single campus; connected supply chains with planned direct freight-rail links and an innovation ecosystem to foster connection and help companies thrive. More...

NEW Webinar: Fundraising – Using LinkSV to Find Investors - Wednesday, July 23 4PM PT

Join Bob Karr and Dawn DeBruyn for a dive into the LinkSV database. With thousands of companies, fundings and investors – VC, corporate, angel – you can find those who have funded companies like yours in the past and may be interested in funding yours now. Plus connecting with the executives in those companies may get you a warm introduction to their investors, and other fundraising tips.

Register here.

Follow us on LinkedIn to stay on top of what's happening in 2025 in startup fundings, M&A and IPOs, VC fundraising plus new executive hires & investor moves.

Early Stage:

• Bedrock Robotics closed a $80M Series A,an autonomous construction technology company that brings proven autonomy systems to heavy equipment fleets.

• Cogent Security closed a $11M Seed, the company pioneering the world’s first AI taskforce for vulnerability management.

• Amber Semiconductor closed a $8.8M Series A, creating a new standard in power management and sensing by enable electricity to be digitally controlled by software in silicon chip architecture.

• WaHa closed a $8M Series A, redefines atmospheric water harvesting, dehumidification, and air conditioning with pure water and dry air at unmatched energy cost.

• Chariot Defense closed a $8M Seed, pioneering next-generation power distribution systems built for the demands of modern warfare that provide modular, high-voltage, and low-signature power at the tactical edge.

Growth Stage:

• Perplexity AI closed a $100M Series E, the builder of the world's first generally available conversational answer engine that draws from credible sources in real-time and directly answers users' questions with sourcing and citations.

• Substack closed a $100M Series C, a new media app that connects you with the creators, ideas, and communities you care about most.

• Exodigo closed a $96M Series B, transforming subsurface intelligence discovery, combining advanced multi-sensing technology and AI to engineer the underground—without digging.

• Unify closed a $40M Series B, an AI-native platform for modern go-to-market teams that combines prospecting data, AI agents, and email sequencing in one unified workflow.

• Fellow Health closed a $24M Series B, a pioneering company dedicated to advancing male reproductive health through innovative and comprehensive mail-in testing solutions.

🌟 Intern Spotlight: Jasmin Garcia 🌟

Your Feedback Matters!

Our mission is to provide an insider's view of Silicon Valley's undercurrents – insights often overlooked by mainstream sources. While many newsletters offer broad market overviews, we focus on delivering a unique, in-depth understanding of the local ecosystem. We share behind-the-scenes conversations, introduce key players we meet at events, and offer exclusive insights.

Your feedback is crucial in helping us refine our content and maintain the newsletter's value for you and your fellow readers. We welcome your suggestions on how we can improve our offering. [email protected] 

Logan Lemery
Head of Content // Team Ignite

Stay up-to-date with AI

The Rundown is the most trusted AI newsletter in the world, with 1,000,000+ readers and exclusive interviews with AI leaders like Mark Zuckerberg, Demis Hassibis, Mustafa Suleyman, and more.

Their expert research team spends all day learning what’s new in AI and talking with industry experts, then distills the most important developments into one free email every morning.

Plus, complete the quiz after signing up and they’ll recommend the best AI tools, guides, and courses – tailored to your needs.

Sign up to start learning.