The Wake-Up Call Nobody Wanted

COVID didn’t just disrupt supply chains. It exposed something worse: we’d been running a confidence game on ourselves for thirty years. When the pandemic hit, 80 percent of active pharmaceutical ingredients flowed through Asia. Eighty percent. That’s not diversification. That’s a single point of failure masquerading as efficiency.

I spent years at McKinsey watching companies optimize their supply chains into fragility. We’d build these beautiful models showing 3 percent cost savings if you consolidated to one supplier, moved manufacturing to the lowest-wage country, and reduced inventory to zero. The math was pristine. The reality was catastrophic the moment anything broke.

The brutal truth is that most companies didn’t learn the lesson until they couldn’t get their products to market. By then, the cost of the crisis exceeded the savings from a decade of optimization.

Friend-Shoring Isn’t Strategy. It’s Risk Management.

Here’s what changed: procurement teams stopped asking “which option costs the least?” and started asking “can we actually execute when things go sideways?” The shift from pure cost optimization to what people call friend-shoring and ally-shoring isn’t ideological. It’s operational realism.

Companies are now explicitly trading some margin for predictability. They’re paying more to source from politically stable countries, allies with aligned interests, and regions with stable labor environments. Is it more expensive? Usually. Is it worth it when your factory doesn’t get nationalized or your shipment doesn’t get stuck in a geopolitical standoff? Absolutely.

Mexico overtaking China as America’s top import source in 2023 for the first time in decades wasn’t an accident. It was the market correcting for concentration risk that nobody wanted to acknowledge five years ago. Proximity matters. Reliability matters more.

The Capital Arms Race Nobody Talks About

While everyone was debating trade policy, governments committed over $100 billion to chip fabrication in the US and EU. CHIPS Act funding, European Chips Act, subsidies for domestic semiconductor manufacturing. This is capital-intensive reshoring that’s reshaping where critical components get made.

The math here is different than it looks. Yes, building fabs domestically costs more per unit. But the cost of supply disruption in semiconductors is existential. Every automotive company, every data center operator, every defense contractor figured this out the hard way. You can’t outsource your bottlenecks.

What’s happening now is selective reshoring of strategic components while non-critical manufacturing stays where economics still make sense. It’s not a return to 1995. It’s a recalibration of what “optimization” actually means when you factor in resilience.

Digital Twins Are Doing Real Work

The operational piece that actually impresses me: digital supply chain twins. These aren’t hypothetical exercises. Companies deploying real-time digital models of their supply chains are reducing disruption response time by 40 percent. That’s not a vanity metric. That’s the difference between managing a problem and drowning in one.

What a digital twin does is let you model scenarios before they happen in the real world. Supplier goes offline? Your model shows the ripple effects in minutes, not weeks. You can test countermeasures, see what breaks, and execute before the actual disruption cascades. That 40 percent improvement in response time is the difference between maintaining service and losing customers.

Check Supply Chain Dive news and you’ll see more companies investing in this capability now. It’s moved from “interesting technology” to “operational necessity” faster than I expected.

ESG Auditing: The Hidden Cost of Not Getting Caught

ESG supply chain auditing is increasing costs. Full stop. You’re paying for auditors, compliance systems, third-party verification, remediation programs. It’s real money. But here’s what it’s actually protecting: your ability to operate.

A decade ago, companies could get away with supply chain blind spots. Now? One exposé about forced labor, environmental destruction, or wage fraud at your supplier and you’re explaining it to shareholders, regulators, and customers simultaneously. The reputational damage often exceeds the cost of preventing the problem in the first place.

ESG auditing is insurance against visibility. It costs money upfront. It prevents much larger costs later. The companies treating it as compliance theater are the same ones taking reputation hits. The ones who actually redesigned their supply chains for transparency aren’t.

What This Means for Operators

The supply chain world that worked from 1990 to 2019 is gone. Cost optimization alone doesn’t win anymore. You need resilience layered in. That means redundancy in critical components, paying more for reliability in strategic areas, and accepting lower margins as the price of staying in business when disruption hits.

It also means the companies that move fastest on this transition have a real competitive advantage. They’ve already absorbed the transition costs and built the operational muscle. They’ll weather the next disruption better than companies still running 1990s playbooks.

For more perspective on how leading operators are reshaping supply chains, McKinsey supply chain insights covers the strategic frameworks companies are using. The data reinforces what we’re seeing on the ground.

What’s your supply chain vulnerability? The companies that win in the next five years will be the ones asking that question hard and building operational answers. If you’re seeing dynamics like these play out in your operations, I’d want to hear about it.

The Moment Everything Changed

In January 2025, a Chinese AI lab called DeepSeek dropped a model called DeepSeek-R1 that did something nobody expected: it matched the reasoning capabilities of OpenAI’s o1 model while costing under $6 million to train. For context, comparable U.S. models cost somewhere north of $100 million. That’s a 95% cost reduction. Not a percentage point reduction. Ninety-five percent.

The market felt it immediately. NVIDIA’s stock fell 17% in a single day—a $600 billion market cap evaporation. That’s the largest single-day wipeout in U.S. stock market history. Investors weren’t panicking about some distant future threat. They were repricing the entire semiconductor advantage that had seemed locked in place. The narrative about moats and defensibility cracked open in 48 hours.

Here’s what matters: this wasn’t some theoretical benchmark game. The DeepSeek-R1 technical report showed legitimate parity on reasoning tasks. And the cost structure made it clear that throwing more money at the problem wasn’t the only lever anymore.

The Price Collapse Nobody Predicted

DeepSeek forced a reckoning. Between January and December 2025, API pricing for frontier AI models dropped 60 to 70 percent across every major provider. Not some models. All of them. OpenAI adjusted. Anthropic adjusted. Everyone adjusted. The pricing power that seemed infinite in 2023 turned out to be quite finite.

This matters operationally, not just philosophically. Your $50,000 annual budget for AI inference today buys you roughly four times what it bought you twelve months ago. That’s not a marginal improvement. That’s a category shift. And it’s still moving.

The cost collapse did something else too: it democratized access. Smaller teams can now afford to run serious AI workflows that required enterprise budgets eighteen months ago. The winner-takes-all dynamics everyone feared never materialized because pricing collapsed before the market could lock in.

What Your Marketing Team Is Already Doing

Marketing teams moved fast. According to the HubSpot 2026 State of Marketing Report, 74 percent of marketing teams now use AI for content generation, up from 48 percent two years ago. That’s not adoption. That’s near-complete migration.

The stated reason? Cost reduction. Not better quality. Not faster time-to-market. Cost. Teams looked at their content budgets, looked at AI pricing, and did the math. The math won.

But here’s where it gets interesting: content generation is the easy layer. It’s table stakes now. The teams winning are pushing AI into harder territory. Personalization engines. Audience segmentation. Predictive lead scoring. These require better models and more compute, but the cost structure now makes them economically viable for mid-market companies.

The Real Opportunity: Customer Acquisition Economics

Forrester Research ran the numbers on AI-driven content personalization for mid-market B2B companies. They found that companies implementing it correctly could reduce customer acquisition costs by 25 to 35 percent within twelve months. That’s not coming from doing less work. It’s coming from doing smarter work with the same or fewer resources.

Think about what that means operationally. If your current CAC is $4,000 and you reduce it by 30 percent, you’re at $2,800. That’s not a rounding error. That’s budget available for scaling or margin you can keep. Most companies would immediately reinvest the savings into volume.

The mechanism is straightforward: AI models are now cheap enough that you can afford to personalize every interaction instead of showing template content. You can segment audiences into cohorts that matter. You can A/B test copy at a velocity that was economically insane two years ago. Volume and precision used to be inversely correlated. Now they’re not.

This requires setup work. It requires thinking about your data infrastructure. It requires some engineering. But the payoff is real and measurable within quarters, not years.

What You Need to Do Right Now

First: audit your current AI spend against pricing today, not last year. You’re probably overpaying for something. Move it. The cost structure is still normalizing.

Second: map where cost reduction alone will move the needle. Content generation is there. Summarization is there. Email copy variations are there. Do these first. They’re easy wins.

Third: get serious about the harder applications. Where does better targeting or faster iteration unlock value? Invest there. This is where 25 to 35 percent CAC reductions come from, not from template generation.

Fourth: build a basic forecast model for your own AI economics. How does variable inference cost change your unit economics? How does it change your growth ceiling? This isn’t MBA strategy work. It’s operational reality. Smart teams are already doing this.

The DeepSeek moment wasn’t just a price shock. It was a repricing of how to think about AI as a cost center versus a growth lever. The cost structure now makes growth math better than it’s been in years. But only if you move fast enough to outrun the rest of your market.

What’s your current roadmap for AI in your team’s operations? I’m genuinely interested in what’s working and what’s becoming table stakes versus differentiator. Drop a note in the comments or reach out.

The Numbers Are Screaming AI Infrastructure, Whether You Want to Hear It or Not

Y Combinator’s Winter 2025 batch just wrapped Demo Day, and if you’re still wondering whether AI infrastructure is a genuine mega-trend or just the latest venture capital infatuation, the data just answered your question. Forty percent of the 170 companies in the batch are explicitly focused on AI infrastructure and developer tooling. That’s not a trend. That’s a market rotation playing out in real time.

Here’s what makes this number matter: it’s a structural shift, not cyclical noise. Back in 2021, when everyone was throwing money at consumer apps and social platforms, you saw maybe 15-20% of a YC batch focused on any single vertical. Now we’re looking at nearly half the entire cohort building the plumbing that powers AI systems. Most market observers get this wrong. The money isn’t chasing the sexiest pitch anymore. It’s chasing what actually solves a real problem at scale, right now.

Valuations Have Come Back to Earth, and That’s Actually Good News

The median pre-money valuation for W25 companies landed around $20 million at Demo Day. That’s down roughly one-third from the $30 million-plus median we saw during the 2021 peak. If you’re a venture capitalist or a founder reading this, you’re probably having two opposite reactions at the same time. It feels like a step backward. It’s also a return to sanity.

Lower valuations mean a few things that matter more than the headline number. First, founders have more room to execute before raising the next round. Second, there’s actual downside protection if a company stumbles during its Series A. Third, and this is the part most people miss, companies are being forced to demonstrate real revenue or real traction before getting deployed at billion-dollar valuations. That’s not pessimism. That’s market discipline finally reasserting itself after three years of what can only be described as valuation theater.

Y Combinator Changed Its Deal, and It Changed the Game Economics

In 2024, YC restructured its standard investment terms. Instead of $125,000 for 7% equity, the accelerator now writes checks for $500,000 at the same equity stake. That’s a 4x increase in capital per percentage point, and it’s not a small adjustment. It fundamentally reshapes the economics for founders exiting the program.

What does this mean in practice? Founders get four times the runway to hit real metrics before they have to fundraise at a Series A valuation. YC gets four times the exposure to winner-take-most dynamics if a company explodes. And the signal to the market is unmistakable: YC is betting on companies that need real infrastructure buildout and real engineering talent, not companies that need to launch and iterate on a landing page. The batch composition confirms this. You don’t deploy an extra $375,000 per company for consumer social apps. You deploy it for deep technical work that takes time and capital to get right.

Defense Tech Isn’t Niche Anymore. It’s Mainstream Accelerator Bets

One of the most telling shifts in W25 is the jump in defense technology and hard tech startups. This category grew faster than any other in the batch, and it’s the thesis Peter Thiel has been pushing for a decade finally breaking into mainstream venture culture. Founders are building satellite communications systems, autonomous logistics networks, and supply chain resilience infrastructure. These aren’t companies that scale to a billion users. They’re companies that scale to billion-dollar budgets.

Why does this matter if you’re not a defense contractor? Because it shows where institutional capital actually believes durability lives. Defense budgets don’t get cut. Government contracts renew. Regulation actually protects your market. Boring compared to chasing the next viral social app, sure, but boring is exactly what sophisticated capital is hunting for in an uncertain macro environment.

The Real Signal: Infrastructure Gets Priced Like Infrastructure Now

Here’s the most important data point that nobody is talking about enough. According to recent venture market analysis, AI infrastructure companies at the seed stage are averaging 18x revenue multiples. Traditional SaaS generalists are trading at 6x. That’s not random pricing. That’s the market saying something very specific: if you’re building the layer that everyone else builds on, you deserve a three-fold valuation premium, even as a seed-stage company.

This is the flip side of lower median valuations. Yes, the overall valuation environment is cooler. But the winners in specific categories are getting priced like they matter. Check the Y Combinator W25 Demo Day coverage and the PitchBook 2026 Venture Monitor if you want the raw data on how these valuations are holding up in the secondary market. Infrastructure companies that solve a specific problem exceptionally well are getting funded at scales that would have seemed insane two years ago, even as the aggregate valuation environment stays grounded.

The W25 batch tells us that startup capital is no longer confused about what it wants to buy. It wants sustainable businesses, defensible positions, real problems solved for real customers. It wants founders who can execute against difficult technical problems, not founders who can pitch better than they can build. This is what a mature startup market looks like. What would you do differently with this information about where capital is actually flowing?

The Unsexy Story That Matters More Than ChatGPT-5

OpenAI just pulled off one of the most consequential corporate maneuvers of the decade, and almost nobody noticed. Late 2024 through early 2025, the company formally transitioned from a capped-profit structure into a full public benefit corporation. This wasn’t a press release moment. It was a messy negotiation with California’s Attorney General, a restructuring that required legal reconfiguration, and a complete reorientation of how the company’s incentives work.

Here’s what matters: This is the governance story every founder operating at scale is either ignoring or misunderstanding. And that’s a problem.

I spent six years at McKinsey watching companies optimize their way into irrelevance. The unsexy operational decisions—the ones nobody gets excited about at board meetings—are usually what determine whether you build something durable or something that implodes under its own contradictions. OpenAI’s restructuring is that kind of decision. It’s not about product. It’s about structure. And structure is destiny.

What Actually Happened, and Why the Timing Matters

OpenAI started as a nonprofit. That was the stated mission: create artificial general intelligence safely for humanity’s benefit. Then reality hit. Building GPT models is obscenely expensive. Training runs cost hundreds of millions. Data infrastructure scales into the billions. The nonprofit structure became a constraint, not a principle.

So the company created a “capped-profit” structure. Microsoft could invest. Employees could get equity. But there were limits on returns, theoretically keeping profit-seeking in check. It was a compromise designed to square a circle: maximize capital while maintaining the original mission.

That compromise broke. During OpenAI’s October 2024 funding round led by Thrive Capital, the company raised $6.6 billion at a $157 billion valuation—the largest venture round in history at that moment. The company needed to move to a traditional structure. Capped profits don’t scale when you’re trying to absorb that much capital.

The formal conversion happened through early 2025 after negotiations with California’s Attorney General. The state had to approve it because technically, OpenAI was still a nonprofit entity with public-serving obligations. The AG essentially signed off on the idea that a for-profit structure would serve the public interest better than a failing capped-profit hybrid.

The Economics Are a Mirage, and That’s the Real Issue

Here’s where this gets uncomfortable. WSJ OpenAI revenue and burn rate analysis showed the company running at $3.7 billion in annualized revenue by mid-2025. That sounds impressive until you look at the burn rate. Analysts were estimating cash burn exceeding $5 billion annually.

Think about that math. The company is making $3.7 billion and spending over $5 billion a year. That’s not a business. That’s a bet.

Now, I get it. Frontier AI requires massive compute investment. You can’t build cutting-edge models on a shoestring. But here’s what matters for founders: OpenAI converted to a for-profit structure while operating at a structural loss. They did this not because they were suddenly convinced that shareholder returns were good for society. They did it because they needed to raise capital, and the capped-profit structure was getting in the way.

The for-profit conversion solved a capital problem. It didn’t solve the economics problem. That distinction matters. The company is betting that future revenue will eventually justify current burn. The board is betting that growth will outpace costs. That’s a standard venture thesis—but it’s being packaged as the natural evolution of the company’s governance, when it’s actually a bet on execution.

Why Elon’s Lawsuit Is Actually About Something Real

Elon Musk sued OpenAI in federal court. The lawsuit, still active as of early 2026, alleges that the for-profit conversion violated the organization’s founding charitable mission. On its face, this looks like a vanity lawsuit from a guy who’s mad about being marginalized from a company he helped start.

But separate from Elon’s motives, the lawsuit raises a substantive question: What does it mean to convert away from a nonprofit mission once you’ve built a constituency around that mission?

OpenAI employees joined a nonprofit with explicit social benefit statements. Early investors bought into the capped-profit structure. The public narrative around the company was always tied to safety-first AI development. The conversion to full for-profit status changes the incentive structure fundamentally. It’s not dishonest. It’s just a different company operating under the same name.

Whether Musk wins the case is secondary. What matters is that the suit is forcing a conversation about governance transition that founders universally underestimate.

The Regulatory Backdrop That Changes Everything

Delaware passed amendments to its Public Benefit Corporation statute in 2025. These amendments, passed specifically in response to high-profile tech governance disputes, introduced new shareholder disclosure requirements that directly affect AI companies restructuring away from nonprofit status.

This is where founder attention needs to be. Delaware amended its corporate law because of exactly what OpenAI did. The state recognized that companies were converting away from benefit corporation status or capped-profit structures without adequate transparency to stakeholders.

The new disclosure requirements mean that if your company is organized as a Delaware PBC and considering a conversion away from that status, you now have explicit obligations to shareholders and the public about that transition. You can’t quietly restructure. You can’t move the mission goalposts without documentation.

For founders scaling a company with explicit social or mission-driven positioning, this matters a lot. Your governance structure is now a regulatory variable, not just an internal choice. The decision to move from nonprofit to for-profit, or to restructure your capitalization, now carries formal disclosure obligations.

What This Means for Your Company

Here’s my actual advice. If you’re a founder building something in regulated territory, operating at scale with multiple stakeholder classes, or anchoring your company to an explicit mission, your governance structure isn’t a checkbox. It’s a strategic variable that will constrain your ability to raise capital, hire, and operate.

OpenAI converted because they needed to. They were operating fine within one structure until that structure became a constraint. That’s sound thinking. But the conversion also happened without full transparency about what was changing and why. The Elon lawsuit exists because stakeholders felt misled.

The lesson isn’t that nonprofit structures are bad or that for-profit structures are bad. The lesson is that you need to make these decisions consciously, with clear-eyed acknowledgment of what changes and what doesn’t. Your stakeholders will hold you accountable to it.

Start thinking about your governance structure now. Don’t wait until you’re at a $157 billion valuation trying to close the largest venture round in history. The economics of your business, the mission you’ve articulated, and the regulatory environment around your industry should all inform this. OpenAI’s PBC transition announcement shows what this looks like when you do it at massive scale. Most companies will handle this earlier and with far less attention. Make sure you’re doing it deliberately.

When $12 Billion Tells You the Market Got Something Right (For Once)

Anthropic just closed a $3.5 billion Series E round. Total funding now sits north of $12 billion. Valuation hit $61 billion. Stop for a second and ask yourself what you actually think that means.

Most people in tech will tell you it’s hype. Another AI arms race. Investors throwing money at anything with “transformers” in the pitch deck. But that’s lazy analysis. That’s what you say when you haven’t looked at where the money actually came from or what it’s being used for.

The real story is different. It’s boring. It’s about infrastructure choices and enterprise procurement criteria that shifted between 2023 and now. And it tells you everything about which AI companies will still matter in five years versus which ones will be fighting for scraps.

The Infrastructure Play That Nobody Calls Out

Here’s what people miss: Amazon committed $4 billion to Anthropic across two separate deals in 2023 and 2024. But the money wasn’t the actual deal. The compute access was.

AWS infrastructure became a structural component of Anthropic’s operations. Not just an option. Not a vendor agreement you can swap out at renewal. A core dependency. That matters more than the headline number because it means Amazon has operational visibility into Anthropic’s scaling trajectory. It means Anthropic’s growth curve is now partially baked into AWS’s capacity planning. Boring? Yes. Strategic? Absolutely.

That kind of embedded relationship is a moat. Not because of the capital injection, but because unwinding it would be organizationally painful. You can find new investors. You can’t easily replace the compute infrastructure running your production models without rebuilding your entire deployment architecture.

The Safety Playbook That Became a Sales Tool

Now here’s where it gets interesting. Anthropic published research on Constitutional AI methodology back in 2022 and updated it again in 2024. The research was technical, rigorous, and honestly kind of dry if you’re not the type who reads model alignment papers for fun.

But something unexpected happened in the market. The EU started using Anthropic’s work as a reference implementation in the AI Act technical annexes. Enterprise IT buyers started asking different questions during vendor evaluations. Safety stopped being something you bolted on and started being something you competed on. According to Andreessen Horowitz’s a16z State of AI 2025 research, 41% of Fortune 500 IT buyers now list safety-as-a-sales-motion as their fastest-growing procurement criterion. That’s not a rounding error. That’s a fundamental shift in how enterprises evaluate AI vendors.

Anthropic built credibility here through actual research and transparency, not through marketing. That credibility is now worth real money in procurement conversations. When your legal and compliance teams start asking whether a model’s alignment methodology has been independently validated, you’re in a different conversation than “is this model 2% faster?” Anthropic gets to answer that question better than anyone else. It’s a moat made of rigor, not hype.

You can read the original methodology in Anthropic Constitutional AI Research. The fact that it’s still being cited by regulators and enterprise security teams two years later is not an accident.

The Benchmark Shift That Changes Everything

Claude 3.5 Sonnet outperformed OpenAI’s GPT-4o on 65% of enterprise coding benchmarks tracked by LMSYS Chatbot Arena in Q3 2024. That number keeps getting buried in tech news cycles. It shouldn’t.

Enterprise procurement isn’t driven by what’s theoretically possible. It’s driven by what solves actual problems. When your development team needs to migrate a codebase or debug production systems, you care about performance on tasks that matter to your business, not abstract benchmarks measuring general reasoning in ways your company will never use.

Sonnet winning on coding tasks means it’s winning on exactly the work that generates the highest volume of AI queries in enterprise environments. That translates to lower costs per successful task completion. Lower costs mean faster payback on AI infrastructure investment. Faster payback means executives approve more deals. That’s the machinery driving market share.

When you combine that technical advantage with Anthropic’s credibility on safety and the infrastructure lock-in from AWS, you get something that starts to look like a defensible position. Not unbeatable. Not permanent. But defensible enough that the $3.5 billion in fresh capital makes sense as a bet that this is real and durable.

What This Actually Means for How Business Works Now

The real moat in AI isn’t going to be raw model performance. It’s going to be the combination of technical rigor, infrastructure depth, and enterprise trust. Those are things you can’t replicate just by spending more money.

Anthropic’s valuation reflects that insight. Not hype, not a bubble, but an actual change in what makes an AI company defensible in the market. The companies that built credibility doing unsexy work, rigorous alignment research, infrastructure partnerships that create mutual dependence, performance optimization on problems that actually pay rent, those are the ones that will still be meaningful in 2030.

Everyone else? They’ll be optimizing metrics and losing market share to competitors who got the strategy right. This is what happens when the hype cycle collides with actual enterprise economics. Boring wins.

Google’s Grip on Search Advertising Is Actually Loosening

Let me start with the number that should make you pay attention: Google’s U.S. search ad market share just dropped to 54.5%. That’s the lowest it’s been since 2008. I know that sounds abstract, so translate it this way. For nearly two decades, Google owned search advertising so thoroughly that the company barely had to innovate on pricing or product. Now there’s an actual alternative pulling share away, and it’s moving faster than most boards realize.

That alternative is Perplexity. The company launched its advertising platform in Q4 2024 and hit 100 million monthly active users by mid-2025. Those numbers matter because they matter to venture capitalists. In January 2025, Perplexity closed a $500 million funding round at an $8 billion valuation led by Institutional Venture Partners. That’s not charity money. That’s institutional conviction that search advertising is genuinely broken and ripe for disruption.

But here’s what most founders miss: this isn’t a story about Perplexity replacing Google. It’s a story about Perplexity fracturing Google’s pricing power in specific, high-value categories. The cracks are already showing.

Where Perplexity Is Actually Winning Against Google

I need to be precise here because the narrative matters. Perplexity isn’t winning across all search advertising categories. It’s winning where it matters most to early-stage startups: B2B SaaS and informational queries.

Early beta advertisers on Perplexity’s platform are reporting cost-per-click rates between $1.20 and $1.80 for B2B SaaS queries. Google’s average across all industries sits at $3.33 per click according to WordStream’s 2025 benchmarks. Do the math. You’re looking at roughly 55-65% lower cost-per-click on Perplexity. That’s not a rounding error. That’s a material arbitrage opportunity.

Why is this happening? Because Perplexity’s user base skews toward people asking complex questions that require synthesis, not just keyword matching. These tend to be higher-intent researchers, and they’re disproportionately early-stage founders, operators, and analysts. They’re not looking for a plumber. They’re looking for a payroll platform or a marketing analytics tool. In B2B SaaS, that user behavior is gold.

Meanwhile, Gartner predicted in 2024 that traditional search engine volume would decline by 25% by 2026 as generative AI tools absorbed informational queries. That’s the core of what’s happening. People are moving informational search queries away from Google’s ten blue links and toward AI-native interfaces. Google isn’t losing brand searches or transactional searches. It’s losing the middle. And the middle is where a lot of B2B advertising happens.

What This Actually Means for Your Ad Budget

Here’s the honest assessment: this creates a window of opportunity for founder-led growth that might not stay open very long.

Perplexity’s pricing advantage exists because the platform has lower overall advertiser competition and a smaller but more targeted user base. That’s sustainable for a while. But the moment growth slows, pricing pressure will evaporate. The platform will need to maximize unit economics. Rates will creep up. Competition will follow. Within two or three years, the arbitrage probably closes.

That means right now, in 2025, founders running B2B SaaS are looking at a genuine pricing window. If you’re currently spending $10,000 a month on Google search ads at $3+ per click, you’re getting roughly 3,000 clicks. On Perplexity at $1.50 per click, you’d get between 6,000 and 8,000 clicks for the same spend. Conversion rates are probably lower because the platform has less advertiser sophistication, but the math still works for many categories.

However. And this is a big however. You can’t just copy-paste your Google campaigns into Perplexity and expect it to work. The user behavior is different. The conversion funnel is different. The platform’s ad interface is less mature. You need to test differently and measure more carefully.

The Bigger Market Shift Nobody’s Talking About

What matters beyond Perplexity is the signal this sends about search advertising more broadly. According to eMarketer Digital Advertising Forecast, Google’s search share erosion is being attributed in part to AI-native search alternatives gaining traction. This isn’t unique to Perplexity. OpenAI is building search features. Microsoft’s Copilot integration is still playing out. Even Reddit and Discord are becoming search destinations for specific intent categories.

What we’re watching is the fragmentation of search advertising. For the better part of twenty years, if you wanted to reach someone in search mode, you went to Google. That’s no longer exclusively true. Now there are parallel search universes, each with different user intent, different pricing, and different advertiser sophistication levels.

For established companies with massive search ad budgets, this is mostly a headache. Google’s still moving the needle at scale. But for startups, especially those with limited ad budgets and high customer acquisition costs, this is genuinely different. You now have multiple channels where informational intent congregates. The math on testing multiple channels simultaneously just got better.

What You Should Actually Do Right Now

If you’re running a B2B SaaS company with a search advertising budget under $50,000 a month, I’d run a structured test on Perplexity. Allocate 10-15% of your current search budget. Build a small set of high-intent keywords tied to your core value props. Measure click volume, conversion rates, and customer acquisition cost. Run this for 60-90 days with discipline. If it works, double down while pricing is favorable. If it doesn’t, you’ve lost a small amount of learning. The downside is capped.

If you’re in a highly competitive SaaS category where Google search ads have gotten expensive, this is less optional. You should test immediately. The pricing arbitrage might be the only way to maintain reasonable CAC.

If you’re spending more than $50,000 a month on Google search ads, this probably isn’t the move yet. Your scale and sophistication advantage is so large that experimenting on an immature platform is likely a distraction. Focus on optimizing what’s already working.

One thing to keep in mind: this pricing opportunity exists because Perplexity is still building. It’s real, but it’s temporary. The next 12-18 months are probably the window. After that, the math normalizes. Founders who move quickly and measure carefully will win. Those who wait will find that window closed.

What’s your current CAC on search ads? Is this shift something you’re already seeing in your own campaigns? I’m curious what you’re actually observing in the market that confirms or contradicts this. Drop your experience in the comments or reach out directly.

The Numbers Stopped Falling. Then They Started Running.

If you’ve been watching venture capital like a stock ticker for the past eighteen months, you already know the feeling. The reset is done. The bloodletting has stopped. What we’re seeing now isn’t recovery from 2023’s crater—it’s something different entirely.

Global venture capital hit $368 billion last year. That’s a 32% jump from 2024. Before you mentally file that under “business as usual,” understand what actually happened: nearly all of that money is chasing one thing. AI deals. Everything else moved sideways.

This matters because it’s not a tide that lifts all boats. It’s a narrowly focused river, and if you’re not in AI, you’re noticing the current pulling away from the dock.

The AI Tier Broke Through Its Previous Ceiling

Series A valuations for AI startups hit $42 million pre-money in Q4 2025. Let that number sit for a second. We haven’t seen that before. The 2021 peak was $38 million. AI companies just took a victory lap around that historical high water mark and kept going.

This is the market saying something specific: we believe the risk-adjusted returns on AI infrastructure, applications, and tooling justify prices we’ve literally never paid for Series A companies before. That’s not enthusiasm. That’s conviction backed by capital allocation.

Here’s the operational reality for founders: if you’re raising an AI Series A in 2026 and your pre-money valuation comes in below $35 million, something’s broken. Either your metrics don’t support it, your team isn’t credible enough, or you’re talking to the wrong investors. The market has reset its floor.

Non-AI Founders Get the Flat-Valuation Treatment

If you’re building fintech, SaaS, logistics software, or anything that doesn’t have “neural” in the pitch deck, your Series A pre-money is still parked at $18 million. Same as Q4 2024. Same as Q3 2024. Flat.

Andreessen Horowitz called this a “two-tier funding market” in their January 2026 market update. That’s partner-speak for “we have one set of rules for AI, another for everyone else.” The candor is refreshing because it validates what every non-AI founder already feels: you’re playing a different game.

The operational implication is harsh but clear. If you’re not AI and you’re raising in 2026, your valuation floor didn’t rise with inflation, revenue growth, or market conditions. It stayed put. You need to either compress your use of capital per dollar raised or accept smaller rounds. Both paths hurt.

Unicorns Are Real Again, But They’re Rarer Than You Think

Eighty-seven new unicorns were created globally in 2025. That’s up from fifty-four in 2024. Progress. Until you remember that 2021 minted 340 of them. We’re at one-quarter the unicorn production rate of peak cycle, yet the number is climbing and the headline sounds optimistic.

Unicorn creation is worth watching because it’s a useful proxy for how generous the top of the market feels. Forty-one more companies crossed the billion-dollar threshold last year than in 2024. That suggests growing confidence. Just not peak-cycle confidence.

Stripe’s private market valuation was pinned at $91.5 billion in a February 2026 employee share transaction. That company was down-rounded to $50 billion in 2023. A $41.5 billion swing in three years. Stripe is the exception that proves the rule—best-in-class execution, undisputed market leadership, and a founder team with permanent credibility. Even for Stripe, the reset was real. It just resolved faster.

Check CB Insights State of Venture 2025 for the granular breakdown on where those eighty-seven unicorns were founded and what sectors they represent. The geographic and sectoral concentration is tighter than the headline suggests.

What This Means When You’re Fundraising

If you’re raising a Series A in 2026, your valuation is now a function of one question: are you AI or not? There’s no middle ground. No “AI-adjacent” or “AI-enabled” hedge. If your core product isn’t powered by large language models, multimodal systems, or AI inference at the product layer, you’re in the non-AI bucket. Your pre-money is $18 million plus or minus 15% based on team strength and metrics.

If you are AI, your floor is higher. But your ceiling is also faster to hit. Investors are pricing in the conviction that your technology works and that you can build. If you can’t show traction, differentiation, or a meaningful moat within months, that conviction reverses quickly.

The broader market data from PitchBook 2025 Annual Venture Monitor shows that capital deployed was driven almost entirely by mega-rounds in AI. The median check size at Series A has actually stayed stable. What changed is selectivity. Capital concentrated. It didn’t democratize.

The Reset Reset

Here’s what you’re actually seeing: the venture market didn’t go back to 2019 or 2018. It went forward into a new structure. AI companies get one set of economics. Everything else gets another. Both are stable now. The turbulence has passed.

That stability is good if you’re in AI or if you’re a best-in-class operator in non-AI sectors. It’s constraining if you’re somewhere in the middle—too expensive to be bootstrap-friendly, not AI enough to attract the concentrated capital.

The valuation reset is officially over. The market is repricing, not recovering. If you’re preparing to fundraise, understand which tier you’re in and plan accordingly. The cost of mistaking one for the other has never been higher.

What’s your read on these numbers? If you’re a founder in market right now, I’d genuinely like to know whether your conversations with investors are matching this data or if something feels different on the ground. Drop a note in the comments or reach out directly.