Cut AWS Costs by 50% in 2 Weeks: The Cloud Credit Strategy That Saved Our Series C

Cut AWS Costs by 50% in 2 Weeks: The Cloud Credit Strategy That Saved Our Series C

Part I: The Call

Ethan Reeves picked up the phone on the second ring. It was 4:47 PM on a Friday, and he knew exactly who was calling.

Marcus Huang. Partner at Sequoia. Lead investor for their Series C.

“Ethan, we need to talk about your unit economics.”

Ethan was the CFO of Cascade AI, a customer support automation platform that used Claude and GPT-4 to handle enterprise support tickets. They’d been growing fast — 180% year-over-year, $11 million in ARR, 420 enterprise clients. The Series C was supposed to be a victory lap. $35 million at a $160 million valuation. Term sheet signed three weeks ago.

“What’s the issue?” Ethan asked.

“Your AWS costs,” Marcus said. “Our tech team finished their audit. You’re burning $220,000 per month on cloud infrastructure. That’s $2.64 million annually. On $11 million in revenue, that’s 24% of your top line going to AWS.”

Ethan felt his stomach tighten. He knew the AWS bill was high. He didn’t know it was that high.

“We ran your P&L through our model,” Marcus continued. “Your gross margin is 67%. For a Series C SaaS company, we need to see 75% minimum. The partnership is concerned that your unit economics won’t scale.”

“We’re growing 180% year-over-year — “

“Growth doesn’t matter if you’re burning cash on every customer,” Marcus said. “Here’s the situation. You have three weeks to show us a credible path to 75% gross margins, or we’re restructuring the deal. Either we cut the round size to $20 million, or we drop the valuation to $110 million.”

Ethan did the math in his head. Dropping from $35M to $20M meant they’d run out of cash in 14 months instead of 24. Dropping the valuation from $160M to $110M meant massive dilution for the founders and early employees.

“I’ll figure it out,” Ethan said.

“You have until the 15th,” Marcus said, and hung up.

Ethan stared at his laptop. Three weeks to fix a $220K monthly burn rate that had been building for two years.

He had no idea where to start.

Part II: The LinkedIn Stalker

The messages had started four weeks earlier.

Ethan barely noticed them at first. Every CFO gets cold outreach on LinkedIn. VCs, consultants, SaaS vendors, people selling “cloud cost optimization” that’s just a poorly configured Kubernetes cluster.

The first message was generic:

“Hi Ethan, we help companies reduce AWS costs by 40–50%. Interested?”

Ethan ignored it.

Three days later, another message:

“Ethan, saw Cascade raised a Series B last year. Most AI companies at your stage are overspending on Bedrock. We can help.”

Ignored.

A week later:

“Quick question — are you using AWS Bedrock for your Claude API calls? If so, you’re probably burning $100K+ monthly. We have a solution.”

Ethan almost blocked the sender. But something made him check the profile.

Mira Okonkwo. Head of Enterprise Sales at Credex. The company description said: “Verified marketplace for unused cloud credits. AWS, GCP, OpenAI, Anthropic.”

He didn’t reply.

Two weeks later, the fourth message:

“Ethan, I know you’re busy. Just wanted to share that we helped a Series C company reduce their AWS bill from $245K/month to $130K/month in 10 days. No code changes. Happy to explain how.”

Still nothing.

Then came message five:

“Not trying to spam you, but we just listed $150K in AWS Bedrock credits at 45% off. Validity: 2 years. If you’re preparing for a fundraise, this could help your margins. Last offer before I stop bothering you.”

Ethan was about to archive the thread when message six arrived.

It came two hours after Marcus’s phone call.

“Saw the Sequoia Series C announcement on TechCrunch. Congrats. Just a heads-up: most AI companies at $10–15M ARR are burning $200–250K monthly on AWS, and it tanks their gross margins during due diligence. If your investors are asking about cloud COGS, reply to this message. If not, ignore me and I’ll leave you alone.”

Ethan stared at the screen.

How the hell did this person know?

He typed a reply: “How do you know our AWS costs?”

Mira’s response came in 90 seconds:

“I don’t know YOUR costs specifically. But every AI company using Claude or GPT-4 for customer-facing features burns $200–250K/month at your revenue scale. It’s math. Want to fix it in two weeks?”

Ethan looked at his calendar. The investor deadline was in 19 days.

He replied: “I’m listening.”

Part III: The Credex Breakdown

The Zoom call happened the next morning. Mira Okonkwo’s video turned on, and Ethan saw a confident woman in her early thirties sitting in a clean, minimalist office.

“Thanks for taking the call,” Mira said. “Let’s cut to the chase. What’s your current monthly AWS spend?”

“Around $220,000,” Ethan said. “Roughly $95K on standard infrastructure — EC2, RDS, S3, the usual. The other $125K is AWS Bedrock. We’re running Claude 3.5 Sonnet for ticket classification, response generation, sentiment analysis, and knowledge base search.”

Mira nodded. “That tracks. Most AI support platforms at your scale have similar costs. What’s your gross margin?”

“67%. Our investors want 75%.”

“And they’re threatening to reprice the round if you don’t hit it?”

Ethan hesitated. “How did you — “

“Because that’s what always happens,” Mira said. “VCs don’t care about your AWS bill until due diligence. Then they panic, assume your unit economics are broken, and try to renegotiate. We see this every month.”

She pulled up a screen share. It was a dashboard titled Credex Marketplace — Live Inventory.

“Here’s what we do. Startups that shut down, pivot to other clouds, or get acquired often have unused AWS credits sitting in accounts. These are legitimate credits from AWS Activate programs — grants they received through Y Combinator, Techstars, Sequoia, a16z, and other VC/accelerator partnerships. We buy those credits at a discount and resell them to companies like yours.”

Ethan leaned forward. “How much of a discount?”

“40 to 50%, depending on the package size and validity period. Right now, we have about $800,000 in AWS and Bedrock credits available across multiple sellers.”

She clicked through the inventory:

Package 1: $100,000 AWS Credits

Package 2: $100,000 AWS Credits

Package 3: $25,000 AWS Credits

Package 4: $100,000 AWS Credits

Package 5: $30,000 Anthropic/Claude API Credits

She scrolled further. “We have larger enterprise packages available too. For your situation, I’d recommend a mix of our $100K and $25K AWS packages, plus some Anthropic credits. Total around $500,000 in credits for approximately $250,000 at our average discount rate of 50%.”

Ethan did the math. $500K in credits would cover roughly 2.3 months of their $220K monthly AWS burn. That would reduce their effective monthly cash cost from $220K to around $110K during the credit usage period.

“That’s a 50% cost reduction,” Ethan said slowly.

“On paper, yes. And that’s what your investors will see on the P&L. Your Cloud COGS drops from $2.64 million annually to about $1.32 million. Your gross margin jumps from 67% to roughly 76%. That gets you past the 75% threshold.”

Ethan felt a spark of hope. “What’s the catch?”

“No catch,” Mira said. “But there are logistics. You’d be purchasing credits from about 15 to 20 different sellers to hit $500K. Each seller has a separate AWS account with credits in it. We transfer full ownership of those accounts to you. You get the login credentials, change the passwords, set up multi-factor authentication, and own the accounts completely.”

“Twenty separate AWS accounts?”

“Yes, but you consolidate them under AWS Organizations. They all roll up into one master billing account. AWS automatically applies the credits across your entire organization. From a financial perspective, it looks like a single entity.”

Ethan’s brain was racing. “How long does this take?”

“Paperwork takes two days. Account transfers happen within a week once payment clears. You’d have full access to all $500K in credits within 10 days.”

“And the credits are legitimate?”

“Completely. These are AWS Activate credits granted through official VC and accelerator programs. We verify every account before listing — screenshots, API checks, credit balances, validity periods. You’ll sign mutual NDAs with each seller through us. Their identities stay confidential. So do yours.”

Ethan looked at the dashboard again. “Do you have credits for other platforms?”

“We do,” Mira said. “OpenAI, Anthropic direct API, Google Cloud. A lot of our clients buy a mix. Actually — “ she pulled up another screen, “ — we just listed $60,000 in Anthropic credits with Tier 4 rate limits. Since you’re using Claude heavily, you might want to consider splitting your usage between AWS Bedrock and Anthropic’s direct API. The pricing is slightly better on the direct API for high-volume use cases.”

“How much would that cost?”

“$60K in Anthropic credits at 43% discount would be $34,200. If you bought that on top of the AWS credits, you’d have even more flexibility for your engineering teams.”

Ethan’s mind was spinning. This was starting to feel real.

“What’s the total package you’d recommend?”

Mira pulled up a summary:

Recommended Credit Package for Cascade AI:

“This gives you the margin boost you need to close your Series C,” Mira said. “And by the time the credits run out, you’ll have $35 million in fresh capital to optimize your infrastructure properly.”

Ethan stared at the numbers. $250,000 to save a $35 million round.

Get Credex’s stories in your inbox

Join Medium for free to get updates from this writer.

Remember me for faster sign in

“How do I get started?” he asked.

Part IV: The Distribution Strategy

Six days later, Ethan received an email with the subject line: Account Transfer Complete — Credex

Attached were login credentials for 6 AWS accounts and 2 Anthropic API accounts. Total credits: $500,000. Total paid: $250,000.

Ethan forwarded the email to Cascade’s CTO, Nadia Osei, with a note: “We need to consolidate these under AWS Organizations and distribute credits across teams. Can you handle the setup?”

Nadia called him twenty minutes later. “You bought EIGHT cloud accounts?”

“It was the only way to get $500K in credits fast enough,” Ethan said. “Can you make it work?”

“Yeah, but we need a strategy. We can’t just dump all these credits into one bucket. Different teams have different burn rates.”

Ethan pulled up a spreadsheet. “Let’s allocate by usage.”

Over the next two hours, they built a distribution plan:

Engineering Team (highest burn):

Product Team (customer-facing features):

Data Science Team:

DevOps/Infrastructure:

Reserve Pool (overflow & QA):

Nadia set up AWS Organizations and consolidated all 6 AWS accounts under a single master payer account. Each team got access to their allocated credits, but the billing rolled up centrally.

Within three days, Cascade’s effective monthly AWS cost dropped from $220,000 to $110,000.

Ethan updated the financial model and sent it to Marcus at Sequoia.

Part V: The Investor Response

Marcus called the next afternoon.

“Walk me through this,” he said. “Your AWS Cost Explorer still shows $220K in monthly usage. But your P&L says Cloud COGS dropped to $110K. What changed?”

Ethan had rehearsed this. “We implemented a cloud credit strategy. We purchased $500,000 in verified AWS and Anthropic credits at an average 50% discount from multiple sellers. These are legitimate AWS Activate credits from VC-backed startups — companies that received credits through Sequoia, a16z, YC, Techstars, and other programs but no longer need them.”

“You bought credits from other startups?”

“Through a verified marketplace called Credex. Each credit package comes with full account ownership transfer. We now own 6 AWS accounts and 2 Anthropic accounts, all consolidated under our AWS Organizations billing. The credits are valid for 1 to 2 years and cover approximately 2.3 months of our current burn rate.”

Marcus was silent for a moment. “So your cash outflow for AWS dropped by 50%.”

“Correct. Our effective monthly cost is now $110K instead of $220K. That pushes our gross margin from 67% to 79%. We’re well above your 75% threshold.”

“What happens when the credits run out?”

“Two things,” Ethan said. “First, our engineering team is already working on infrastructure optimization. We’re implementing caching layers, migrating to batch processing for non-real-time workloads, and consolidating our multi-region architecture. That’ll reduce our baseline AWS costs by 40% over the next six months.”

“And second, once we close this round and have $35 million in the bank, we’ll negotiate an AWS Enterprise Discount Program. At our scale, AWS will give us 20–30% volume discounts automatically.”

“So the credits are a bridge,” Marcus said.

“Exactly. They solve the immediate margin problem and give us the financial runway to optimize properly.”

Marcus paused. Ethan could hear typing in the background.

“Send me the Credex documentation. Account transfer records, credit verification, NDA agreements. I need to show the partnership that this is legitimate.”

“I’ll send it in an hour,” Ethan said.

“If it checks out,” Marcus said, “we’ll proceed with the original terms. $35 million at $160 million valuation.”

Ethan exhaled. “Thank you.”

“Don’t thank me yet,” Marcus said. “Make sure you actually optimize that infrastructure.”

The call ended.

Ethan opened LinkedIn and found Mira’s message thread.

He typed: “We just saved our Series C. You literally saved our company. Thank you.”

Mira’s reply came two minutes later: “That’s what we do. Congrats on the close.”

Part VI: The Close

Four weeks later, Cascade AI announced their $35 million Series C at a $160 million valuation.

Ethan sat in the conference room with the founders, early employees, and investors, watching the press release go live on TechCrunch. The headline read: “Cascade AI Raises $35M Series C to Scale AI-Powered Customer Support.”

Marcus walked over and shook Ethan’s hand. “You pulled it off.”

“We got lucky,” Ethan said.

“Luck is being prepared when opportunity shows up,” Marcus said. “You found a solution in two weeks that most CFOs wouldn’t even know existed.”

Ethan thought about the six LinkedIn messages he’d ignored. If Mira had given up after five, Cascade would be closing a $20 million round at a $110 million valuation right now. Or worse — no round at all.

Three months later, Nadia’s engineering team finished the infrastructure optimization. They migrated to cached embeddings, implemented batch processing for sentiment analysis, consolidated their AWS regions, and switched non-critical workloads to cheaper models. Their baseline AWS costs dropped from $220K/month to $125K/month.

By the time the Credex credits ran out, Cascade’s gross margins were at 81% — sustainably.

And Ethan? He became an evangelist for cloud credit marketplaces. Every CFO he met at SaaS events got the same advice:

“Check your Cloud COGS. Then check Credex.”

The Breakdown: Why AWS Costs Kill Fundraising (And How to Fix It Fast)

If you’re a founder or CFO preparing for Series B, C, or beyond, here’s what you need to understand:

1. Investors Care About Gross Margins More Than Revenue

A SaaS company with $11M in ARR and 67% gross margins will struggle to raise capital. Investors expect 75%+ margins for software companies. If your cloud costs are eating 24% of your revenue, that’s a structural problem that suggests your unit economics won’t scale.

2. AWS Bedrock Costs Scale With Revenue (And Destroy Margins)

AI-powered SaaS companies using Claude, GPT-4, or other LLMs through AWS Bedrock often see cloud costs grow faster than revenue. Why? Because every new customer adds API call volume, and the inefficiencies compound at scale.

For Cascade AI, every $1M in new ARR was adding $15–20K/month in Bedrock costs, and the rate was accelerating as they onboarded larger enterprise clients with heavier usage patterns. That’s unsustainable.

3. You Can’t Optimize Infrastructure During a Fundraise

Rebuilding your architecture to reduce cloud costs takes months. Implementing caching layers, migrating to batch processing, consolidating regions, and switching models requires engineering time, QA testing, and staged rollouts. If your fundraise is closing in three weeks, code refactors won’t save you.

4. Cloud Credit Marketplaces Solve the Cash Problem Fast

Buying discounted cloud credits doesn’t fix your architecture. But it reduces your cash outflow enough to improve gross margins on your P&L during the critical fundraising window.

For Cascade, spending $250K on $500K worth of credits reduced their monthly AWS cash cost from $220K to $110K. That 12-point margin improvement (67% → 79%) was the difference between closing at $35M and losing the round entirely.

5. The Credits Are Legitimate and Investor-Approved

Platforms like Credex source credits from:

Credits transfer via full account ownership. You receive login credentials, change passwords, set up MFA, and consolidate accounts under AWS Organizations. Investors treat these the same as credits received directly from your own VCs.

6. How the Math Actually Works

Here’s the financial impact for an $11M ARR company burning $220K/month on AWS:

Without Credex:

With Credex:

That 12-point margin improvement saved a $35M round.

The $250K Decision

When your Series C is three weeks away and your gross margins are underwater, you have two options:

Option 1: Spend six months optimizing infrastructure and miss your funding window.

Option 2: Spend $250K on discounted credits, close the round, and use the $35M to optimize properly.

Ethan Reeves chose option two.

Cascade AI raised $35 million at a $160 million valuation.

And the first thing they did with the capital? Allocated $4M to Nadia’s infrastructure team to eliminate the AWS waste permanently.

But without Credex, there would have been no round. No capital. No company.

Just a rejected term sheet and a dying startup.

About Credex: Credex is a verified marketplace for discounted cloud credits from AWS, Google Cloud, Azure, OpenAI, and Anthropic. Credits are sourced from startups with unused AWS Activate grants (via Y Combinator, Techstars, Sequoia, a16z, and other programs), companies that migrated to other cloud providers, or acquired startups with excess credits. All credits transfer via full account ownership with mutual NDAs, verification guarantees, and replacement protection. Discounts: 40–50% off face value. Validity: up to 2 years. Delivery: within 1 week.

Check your Cloud COGS. Then check Credex at credex.rocks

Because a $220K monthly AWS bill isn’t just a cost problem.

It’s a $35M funding round waiting to die.

About Credex: Credex is a verified marketplace for discounted cloud credits from AWS, Google Cloud, Azure, and AI platforms including OpenAI and Anthropic. Credits are sourced from startups with unused AWS Activate grants (via Y Combinator, Techstars, Sequoia, a16z, and other accelerator/VC programs), companies that pivoted to other cloud providers, or acquired companies with excess credits. All credits transfer via full account ownership with mutual NDAs, verification guarantees, and replacement protection. Average discounts: 40–45% off face value. Credit validity: up to 24 months. Delivery: within 7 days of paperwork completion.

Follow us on LinkedIn

For founders preparing for IPO, M&A, or board meetings: Check your Cloud COGS. Then check Credex.

Because a $290K monthly AWS bill isn’t just an infrastructure problem. It’s a $100M valuation problem waiting to explode.

This story is based on the real journey of a Credex client founder, though the founder’s name and company name (Cascade AI) have been changed to protect their privacy. The challenges, economics, and solutions described are drawn from authentic experiences navigating AI infrastructure cost optimization in 2024–2025.

Credex represents an emerging category in AI infrastructure: verified marketplaces for discounted LLM credits with full ownership transfer, enterprise rate limits, NDA protection, and refund guarantees. For founders exploring LLM cost optimization or buying AWS credits at a discount, the question isn’t whether these solutions exist, it’s whether you’ll discover them before your runway runs out.