Startup Funding Watch: Smart Capital Is Barbelling Between Cost-Control Software and Physical Infrastructure

The CODEW Startup Funding Watch | August 14, 2026

This week's capital flows split into a clean barbell: money piling into the software layer that helps enterprises control AI costs, and money piling into the physical infrastructure that makes AI possible at all. Almost nothing landed in between. Here's where smart capital is actually placing its bets.

THE FUNDING LEAD

Databricks Wanted $1 Billion. Investors Offered $15 Billion. It Settled at $5 Billion — and a $190 Billion Valuation.

Amount / Valuation: $5 billion at a $190 billion post-money valuation, up from $134 billion seven months ago. Investors: Coatue (lead), Blackstone, MGX, T. Rowe Price, and new investor Sixth Street Growth, alongside returning backers Andreessen Horowitz, Thrive Capital, Goldman Sachs Alternatives, and Temasek. Business: Enterprise data-and-AI platform, with proceeds directed at Unity AI Gateway (multi-model cost routing), Lakebase (agent-native database), and Genie (business AI agent). Use of capital: scaling the orchestration tools enterprises use to manage spend across multiple AI vendors.

Why the capital is moving here: CEO Ali Ghodsi told TechCrunch the company only wanted $1 billion — a report leaking mid-raise during Databricks' own June conference triggered an unsolicited bidding war that investors pushed toward $15 billion before settling at $5 billion. That's not hype chasing a hot sector; it's investors actively competing to get into a company positioned as the toll booth between enterprises and an increasingly fragmented, multi-vendor AI model market. Databricks' own AI Gateway has now routed more than one quadrillion token, giving it a real-time read on a shift Ghodsi calls "tokenmaxxing to valuemaxxing" — enterprises moving from consuming AI tokens indiscriminately to demanding measurable value per dollar spent.

What it reveals: The bet isn't on Databricks having the best model — it doesn't have one. It's on Databricks being the layer enterprises trust to make sense of everyone else's models as costs climb and vendor choice multiplies. That's a fundamentally different, and in this case more oversubscribed, thesis than backing a model lab directly.

AI Funding Watch

Capital chasing AI-native coding tools is re-rating faster than almost anywhere else in venture right now. Cognition, maker of the Devin coding agent, is in early talks to raise $1 billion-plus at a valuation above $40 billion — up from $26 billion just three months ago in May. Lovable, the Stockholm-based "vibe coding" platform, confirmed a $400 million Series C at a $13.3 billion valuation, roughly doubling its $6.6 billion mark from December, led by Menlo Ventures and the Scaleup Europe Fund; the company says it crossed $500 million in annualized run-rate revenue in June. And Blacksmith, an AI code-testing and validation startup, raised a $45 million Series B led by Peak XV Partners at a $550 million valuation — a roughly tenfold jump from the $60 million valuation it held after its 2025 Series A.

The pattern across all three: capital is moving specifically toward tools that make AI-generated code trustworthy and deployable at scale — writing code, testing it, and validating it — rather than toward foundation models themselves. That's capital flowing toward the application and workflow layer, not the model layer, and it's happening at a pace (a company more than doubling valuation in eight months, another jumping tenfold in about a year) that's genuinely unusual even by this year's standards.

Sector Capital

Deep tech and energy continue absorbing mega-rounds most other software categories can't command: Valar Atomics raised $1 billion in a Series B for modular nuclear reactors built specifically to power AI data centers, led by Sequoia Capital with a $200 million JPMorgan-led credit facility attached; Base Power raised $1 billion in a Series D for battery systems designed to absorb AI-driven grid demand spikes. Both closed in the same week earlier this month, and both reflect the same underlying bet — that AI's power requirements are themselves investable infrastructure, not just a cost line for hyperscalers to absorb.

AI infrastructure hardware is seeing similar conviction at the chip layer: London-based OLIX Computing raised $312 million at a $3.3 billion valuation for photonic AI inference chips, a genuine architectural bet against electronic inference as the permanent standard.

Cybersecurity remains anchored by Cyera's $600 million round at a $12 billion valuation, led by Evolution Equity Partners — still the largest AI-security check of the year and the benchmark every subsequent round in the category gets measured against, particularly as enterprises scramble to secure AI agents with real permissions inside their systems.

Robotics capital is flowing more unevenly. Physical Intelligence, building general-purpose robot foundation models, is reportedly in talks for a fresh round approaching $1 billion at roughly $11 billion — nearly double its valuation from four months earlier, even before commercial deployment at scale. That continues a run that saw Skild AI triple to a $14 billion valuation on a SoftBank-led round earlier this year, with Nvidia's venture arm participating — evidence that robotics investors are underwriting team pedigree and technical promise well ahead of revenue, the same pattern frontier AI labs saw in their earliest funding rounds.

Investor Watch

Coatue is having a genuinely broad quarter, leading Databricks' $5 billion round while continuing to back physical-AI bets elsewhere in its portfolio — a signal the firm sees software orchestration and physical AI infrastructure as complementary theses rather than competing ones. 

Menlo Ventures leading Lovable's Series C extends its recent pattern of high-conviction bets on AI-native developer tools. Sequoia Capital's lead on Valar Atomics, paired with a bank-led credit facility rather than pure equity, shows growth-stage venture increasingly blending debt and equity to fund capital-intensive hardware bets — a structure more common in project finance than traditional startup investing. And SoftBank's continued heavy robotics posture, visible in both Skild AI's round and its broader 2026 activity, marks it as the most consistent large-check believer in humanoid and general-purpose robotics specifically, ahead of most traditional venture firms.

Valuation & Market Signals

The valuation data this week shows two distinct re-rating speeds coexisting in the same market. AI-native application startups are repricing explosively: Lovable doubled in eight months, Blacksmith rose roughly tenfold in about a year, and Cognition's reported talks would mark a 54% step-up in three months alone. 

Databricks, despite being the most oversubscribed round of the week — investors pushed for three times what the company asked for — still priced at a comparatively measured 42% step-up over seven months. That gap is worth sitting with even the fastest-growing, most fought-over infrastructure company in this week's data is being priced more conservatively than pure AI-application startups, suggesting investors still apply more discipline to infrastructure valuations than to app-layer ones, even when infrastructure demand is unambiguously stronger.

Startup to Watch

Lovable. A Stockholm-based "vibe coding" platform doubling its valuation in eight months on the back of tripled ARR is a genuinely unusual trajectory for a European AI startup, in a category historically dominated by U.S. names like Cursor and Replit. Lovable's $500 million run-rate revenue as of June suggests real usage behind the valuation, not just narrative — and its position squarely in the AI-native coding cluster means its next few quarters will be a useful bellwether for whether this category's re-rating pace is sustainable or a sign of overheating. Worth watching specifically for whether its growth rate holds as Cognition, GitHub Copilot, and Cursor all compete more aggressively for the same non-technical "vibe coding" audience Lovable has built its early lead on.

Capital & Competition

Databricks' raise sharpens its position directly against Snowflake, its closest publicly traded rival in enterprise data platforms — a $190 billion private valuation now dwarfs Snowflake's public market cap, giving Databricks both a war chest and a valuation umbrella Snowflake's shareholders will be watching closely. Ghodsi's comment that Databricks is "very unlikely" to IPO before either Anthropic or OpenAI is itself a competitive signal: it suggests Databricks sees its own public listing as contingent on watching how the market absorbs the two AI labs' eventual debuts first, rather than racing to be first. 

In coding AI, Cognition, Lovable, and Blacksmith's near-simultaneous re-ratings collectively raise the competitive bar for GitHub Copilot and other incumbent developer tools — three well-funded, fast-growing challengers converging on overlapping workflows in the same few months is the kind of capital concentration that tends to accelerate consolidation, either through M&A or through one player's growth rate pulling decisively ahead.

Three Funding Signals

1. Capital Is Rewarding the Orchestration Layer Over Raw Model Access

Databricks' oversubscription confirms investors see more durable value in managing multi-vendor AI complexity than in backing any single model provider — a thesis likely to pull more capital toward cost-control and routing tools over the next 6–18 months.

2. AI-Native Coding Tools Are the Fastest-Repricing Category in Venture Right Now

Three separate startups doubling-to-tenfolding in months signals genuine category conviction but also concentrates real correction risk if growth rates in even one of them disappoint a market pricing in continued acceleration.

3. Physical AI Infrastructure Is Absorbing Capital as Fast as Software, Not Instead of It

Nuclear power, battery storage, photonic chips, and robotics all pulled nine- and ten-figure rounds in the same window as this week's software mega-deals — investors are hedging across both ends of the AI stack simultaneously rather than picking one layer to bet on.

THE CODEW TAKE

Smart capital this week is building a barbell, not a bet on a single winner. On one end, Databricks' oversubscribed round shows investors paying up for the software layer that helps enterprises manage an increasingly commoditized, multi-vendor AI model market — the orchestration and cost-control tools that get more valuable, not less, as models themselves get cheaper. On the other end, nuclear reactors, grid batteries, photonic chips, and robot foundation models are absorbing capital at a pace that shows no sign of slowing, because none of the software layer matters if the physical capacity to run AI at scale doesn't exist. 

What's conspicuously thin in between is capital chasing generic AI applications without a clear cost, infrastructure, or workflow thesis attached — the middle of the stack, not the software or hardware extremes, is where this week's investors showed the least conviction. The technology market that capital is trying to build isn't a single winner-take-all platform; it's two distinct, well-funded stacks — one managing AI's economics, one building its physical capacity — that will increasingly determine which AI companies can actually scale.

---

Source Attribution

1. CNBC — Databricks Funding Round Hits $190 Billion Valuation
2. TechCrunch — Databricks Wanted to Raise $1B, Investors Wanted $15B. It Settled on $5B at a $190B Valuation
3. Forbes — Databricks Hits $190 Billion Valuation As CEO Ali Ghodsi Claims AGI Already Arrived
4. Bloomberg — AI Startup Cognition in New Funding Talks at $40 Billion Value
5. TechCrunch — Lovable Confirms New $13.3B Valuation, Raises Another $400M
6. Wall Street Journal — Vibe-Coding Startup Lovable Hits $13 Billion Valuation
7. TechCrunch — Blacksmith's Valuation Jumps 10x to $550M as AI Coding Fuels Software Validation
8. Tech Startups — Venture Capital & Startup Funding Roundup, August 4, 2026
9. Crunchbase News — So Far, 2026 Is a Solid Year for Cybersecurity Startup Funding (Cyera)
10. Bloomberg — Ex-DeepMind Staffers' Robotics Startup in Talks for $11 Billion Valuation (Physical Intelligence)
11. Crunchbase News — Robotics Startup Skild AI Lands $1.4B, Tripling Valuation to $14B in Just 7 Months

THE CODEW · STARTUP FUNDING WATCH

Editorial Note

The CODEW Startup Funding Watch tracks venture capital, startup financing, emerging technology companies, investor activity, and the sectors attracting strategic capital across the technology industry.

Funding information is based on company announcements, investor disclosures, public filings, industry reporting, and other publicly available sources. Reported funding amounts and valuations may reflect different transaction structures and reporting periods.