Drowning in Batteries

How Powin went from market leader to bankrupt

Author

Nicolás Cerdeira
June 19, 2025

Hey — It’s Nico.

Welcome to another Failory edition. This issue takes 5 minutes to read.

If you only have one, here are the 5 most important things:

This issue is brought to you by dofollow.com — the link and citation building agency that helps your startup rank on ChatGPT, Perplexity, and Gemini.

Let’s get into it.

In 2025, AI ranks the winners—smart founders invest in visibility.AD

AI and search are how your buyers discover products.

To show up, your SaaS brand needs authority. That means backlinks.

dofollow.com helps founders earn editorial links from high-authority, niche-relevant sites—fast.

Just ask Heymarket, who scaled from 4K to 48K+ monthly visitors without adding headcount.

We’ve done the same for brands like Surfshark, Pitch, and Wiz.

Here’s what you get:

→ Backlinks from high-authority, niche-relevant sites

→ Real outreach in your vertical, no mass lists

→ Monitoring + free replacements if a link goes down

→ No long-term contracts—just outcomes

This Week In Startups

🔗 Resources

An engineer's guide to vibe design (with prompts).

Vertical AI & the Productivity Paradox

📰 News

Midjourney launches its first AI video generation model, V1.

Own, a new social media app, aims to tokenize the creator economy.

💸 Fundraising

AndrenaM, a startup building AI-powered sonar tech to "secure the oceans", raised $10M.

Vehicle intelligence startup Applied Intuition raises $600M

Defense technology startup Mach Industries raises $100M

Tennr raises $101 million to fix the patient referral process.

Fail(St)ory

Power Cut

This week, Powin, a U.S. battery manufacturer once hailed as a major player in the clean energy transition, filed for bankruptcy.

Just last year, they were one of the biggest players in the U.S. market for grid-scale batteries. Now they’re sitting on over $300 million in debt, most of their team is gone, and the company is over.

What Was Powin:

Powin built utility-scale battery systems—the kind used by power companies to store energy and stabilize the grid.

Its main product, the Powin Pod, was a 20-foot container designed to store and release energy on demand for utilities—think of it as a massive battery bank for the grid.

The pitch was speed and simplicity. Powin’s containers arrived pre-wired and ready to deploy. Utilities could plug them in, stack them side by side, and scale fast. Everything came integrated: the batteries, the thermal management, the safety systems, and a software platform—StackOS—that let operators monitor performance in real time and respond to energy demand.

Their customers were mostly utility companies and energy developers looking to build large-scale storage projects, especially to back up solar and wind. Powin wasn’t reinventing battery chemistry, they were focused on making deployment easier, faster, and more reliable at scale.

And for a while, it worked. Their systems were deployed across the U.S. and internationally. By 2023, Powin had become the third-largest energy storage provider in the U.S. by installed capacity. Their growth tracked closely with the clean energy boom—and they were seen as one of the few American companies capable of competing with Chinese battery giants.

They weren’t a hype-driven startup. Powin had been around for over a decade, quietly evolving from a small energy firm into a serious clean tech player. That’s what makes the collapse so surprising.

The Numbers:

  • 📅 Founded: over a decade ago, taken private in 2018.

  • 💸 Raised: $360M in multiple rounds.

  • 💼 Team: Had over 400 employees at its peak.

  • 💸 Declared bankruptcy: June 2025, with $300M+ in debt.

Reasons for Failure: 

  • Tariffs Crushed Their Supply chain: They built their business on cheap Chinese battery cells and didn’t have a backup plan. When tariffs on Chinese lithium-iron-phosphate (LFP) cells spiked, Powin’s costs went through the roof. They tried to shift to U.S. suppliers, but the local market wasn’t ready to deliver at scale. In a low-margin industry like this, that kind of hit is hard to absorb—and they couldn’t.

  • They couldn’t pay their suppliers: Powin owed tens of millions to key partners like CATL and Qingdao CIMC Prowin—two major players in the battery supply chain. CATL even took legal action to recover $44 million in unpaid shipments. When your suppliers are suing you and showing up on your bankruptcy creditor list, it’s a sign the business was already breaking down behind the scenes.

  • Lack of Vertical Integration: Unlike players like Tesla who build their own cells, Powin was mostly a systems integrator. That gave them speed, but also meant they had little control over their most important input—batteries. As soon as their suppliers got expensive or unreliable, Powin had no leverage.

  • Too much debt, not enough time: Powin raised $135M from investors—but then borrowed another $200M to keep growing. That’s a lot of pressure in a business where deals move slow and profits are thin. When costs jumped and projects stalled, the debt piled up fast. They weren’t just running out of money—they were running out of runway.

Why It Matters: 

  • Not even market leaders are safe when their supply chain isn’t under control.

  • Fast growth backed by debt only works if your costs are predictable

  • In low-margin industries, one external shock can kill even the top players.

Trend

YC’s Spring Demo Day

This week, Y Combinator held its first-ever Spring Demo Day, showcasing 144 startups that just wrapped the new April–June cohort. And if you’re wondering whether YC has moved on from AI agents — it hasn’t.

Out of 144 companies, nearly half were building agentic AI in one form or another. It wasn’t just a theme. It was the cohort. And while many were predictable wrappers around ChatGPT, a handful pointed to something more interesting: actual patterns.

This is not a story about AI agents being “the future.” That’s old news. This is about what that future is starting to look like — which ideas are sticking, which problems founders are choosing to solve, and how YC is quietly shaping the next layer of the AI stack.

Why It Matters:

  • The agent hype is evolving into structure. We’re seeing real categories emerge: vertical tools, infrastructure, workflows.

  • YC isn’t just amplifying AI — it’s steering it. The batch reflects what they want to fund, and what the next wave might build on.

  • These agent startups give us a preview of what the next SaaS wave might look like — leaner, more specialized, and powered by orchestration, not UIs.

The Trends

1. Vertical AI Copilots

The loudest trend in the room was verticalization — AI agents built for one job, in one industry, with domain-specific knowledge baked in. These aren’t generic chatbots. They’re process automation tools dressed as copilots, replacing humans in narrow, painful workflows.

  • SynthioLabs is building an AI-powered sales rep for pharma companies — one that actually knows medicine. Instead of spamming doctors with generic emails, it crafts personalized outreach that’s clinically accurate and always on. It’s not just sales automation — it’s a medical rep that never sleeps.

  • Eloquent AI automates customer operations in financial services. You want to unfreeze your bank account or add a driver to your policy? Their bots do it instantly — no engineering team needed.

  • Approval AI does the same for mortgages. One-click mortgage: the agent shops rates, negotiates, and fills out your paperwork.

  • Beluga Labs wants to be Rippling for content creators, helping them automate taxes and analyze income streams.

This trend isn’t new, but it’s deepening. These startups are betting that domain context beats general intelligence. And YC seems to agree.

2. Agents for the Rest of Us

Another theme: accessibility. A bunch of teams are trying to drag AI agents out of the browser tab and into your daily life — chats, voice, even tools non-technical users can build themselves.

  • Text.ai is embedding agents inside SMS, WhatsApp, and Telegram, making AI available without an app store, login, or tutorial. Just text and go.

  • Willow adds voice dictation to everyday messaging, using an agent to rewrite what you say in your tone.

  • Sim Studios wants to make building agents as easy as dragging blocks in Figma. Their open-source builder is already being used by the U.S. Department of Defense and some big-name customers.

This wave is less about breakthroughs and more about reach. It’s about pushing agents into familiar channels, making them feel ambient. The idea: if you want mass adoption, don’t teach people new behaviors — meet them where they already are.

3. Infrastructure for the Agent Economy

Agents don’t just need prompts — they need tooling, guardrails, and observability. Behind the scenes, YC backed several startups building the scaffolding to help others build and scale AI agents safely and effectively.

  • LLM Data Company evaluates agent performance using its own internal models. Turns out, testing whether an agent is “good” is a hard problem — and one Perplexity is already paying them to solve.

  • Anvil is trying to redefine SEO for LLMs. As search shifts from keywords to prompts, Anvil helps brands get surfaced in AI-native interfaces like ChatGPT and Gemini.

  • Casco stress-tests agents by simulating attacks. Basically, red-teaming-as-a-service for autonomous workflows.

4. AI Eating SaaS from the Inside

The last big trend: full-stack replacements for knowledge work tools. Not integrations. Not add-ons. These startups want to replace Slack, Notion, SAP, and your analyst — all in one go.

  • Den is pitching itself as a “Cursor” for enterprise knowledge — a fully agent-powered hub where employees interact with internal tools through conversations.

  • Auctor automates enterprise software implementation, and claims even vendors like AWS and SAP have reached out to use it for their own onboarding flows.

  • Scalar Field wants to be the AI Bloomberg terminal — not just a dashboard, but an agent that manipulates financial data and surfaces insights.

  • Airweave turns your existing productivity tools into searchable agent-ready databases — kind of like turning your internal chaos into ChatGPT fuel.

Help Me Improve Failory

How Was Today's Newsletter?

If this issue was a startup, how would you rate it?

Login or Subscribe to participate in polls.

That's all of this edition.

Cheers,

Nico