Accelevation IPO valuation drops from $5.4B target after $18 pricing

From a $5.4 billion target to a below-range pricing

In late September, Accelevation set out to become a public test of how much investors would pay for the physical equipment behind AI data centers. The company makes and installs power and cooling systems for data centers. When it launched its roadshow on September 22, it said it was aiming for a valuation of up to $5.37 billion in a US initial public offering, and it planned to list on Nasdaq under the ticker "ACCV".11 The company and some selling shareholders offered 30 million shares at $20 to $24 each. At the top of that range the deal would have raised as much as $720 million.15

The deal did not price at those terms. On September 29, Accelevation and its backers sold the same 30 million shares at $18. That was $2 below the bottom of the range and raised $540 million.5 When trading opened on September 30, the stock started at $17.55, 2.5% below the offer price, which valued the company at $3.92 billion.8 About a month later, the shares were trading well below the IPO price. They closed at about $14.48 on October 30.23

The change in the composition of the offering also matters. In my view, it is the most revealing part of the deal and has received less attention than the headline price.

Who was actually selling

The original filing split the 30 million shares into 8.64 million new shares from the company and 21.36 million existing shares from current holders. That put roughly 71% of the base offering in the hands of selling shareholders.16 The main seller was Olympus Partners, the private equity firm that bought Accelevation in January 2025.1920 The underwriters' option to buy another 4.5 million shares also came entirely from the selling stockholders.19

One critical analysis of the filing put it simply: at the top of the range, the company would have kept only about 29 cents of every dollar raised, and the rest would have gone to Olympus and its co-investors.20 Reporting that sorted the dollar amounts by seller reached the same split. About $513 million would have gone to existing holders and roughly $207 million to Accelevation.16

The final deal shifted that balance. At $18, the company sold 10 million shares and Olympus-affiliated holders sold 20 million.410 That works out to about $180 million for Accelevation and $360 million for Olympus.6 Accelevation's own net proceeds came to about $170.6 million, which it said would go toward repaying debt.4 Even after selling more shares, Olympus was expected to keep roughly 85% of combined voting power.1 A later ownership table showed Olympus Advisors holding about 73% of the shares.22

So the company issued more new stock at a lower price while the private equity owner sold fewer shares. That looks like a negotiated compromise to get the deal done. The sources do not describe the talks between bankers and buyers, so this is my reading of the numbers rather than a reported fact.

A real business with thin margins and heavy debt

The business behind the offering is not in dispute. Accelevation was founded in 2017 by Michael and Shawn Rubiera, and revenue grew from less than $3 million in 2021 to $447.8 million in 2025.8 Revenue for the 12 months ended June 30, 2026 was $727 million, and the backlog at that date was about $1.1 billion.3 Its book-to-bill ratio was 3.5x for the quarter ended June 30. It runs about 1.1 million square feet of US manufacturing capacity and has about 1,716 employees, nearly half of them in field services.4 Its customers reportedly include Google, Amazon, Microsoft and Meta.6

Writers who were enthusiastic about the deal focused on those numbers. One venture-focused analysis called it the most underrated IPO of the fall for investors betting on AI infrastructure. It argued that power distribution is the unglamorous layer of the AI buildout that most investors overlook.9

The skeptics looked at the other half of the filing. Net income was $21.8 million on $448 million of revenue, a net margin under 5%. The margin slipped to 4.4% in the first half of 2026 even as sales rose.20 Debt went from $72.7 million to about $648 million after the 2025 buyout.20 As of June 30, about $651.5 million was outstanding under the company's credit facility.6 At the top of the range, the same critical analysis put the valuation at roughly 250 times trailing earnings. It compared that with about 56 times for Vertiv and 45 times for Eaton, both of which are far more profitable.20 Later coverage pointed to higher labor costs and project execution problems as pressures on gross margin.25

I think the skeptics read the risk correctly, but they framed it a little too narrowly. Accelevation is not a speculative story stock. It has contracted demand and a growth rate few industrial companies can match. The problem was that it went to market priced like a high-margin franchise while the financials looked like a fast-growing, low-margin manufacturing and installation business. Because so many shares came from the seller, a large share of the cash raised was never going to strengthen the company's balance sheet.

Market conditions

Company-specific concerns were only part of the story. Much of the coverage put more weight on the broader market, and I largely agree. Reuters noted that the fall IPO season began hesitantly, amid surging bond yields and the first US interest-rate hike in three years.13 When the deal priced, coverage linked the discount to higher yields, which make investors demand cheaper prices for companies whose biggest profits are still in the future.12

The wider AI infrastructure trade was also under strain. Reuters cited the reported delay of SB Energy's IPO and a dispute involving Oracle and Blue Owl that could slow a data center project in New Mexico.8 Matt Kennedy, a strategist at Renaissance Capital, said market conditions had worsened and that new AI stocks were no longer a sure thing. He pointed to Holtec Nuclear's postponed listing as another sign of strain.7

Kat Liu of IPOX predicted the shift before pricing. She said investors would stop treating AI infrastructure as one hot theme and would look more closely at a company's place in the value chain, the quality of its backlog and how well it turns demand into margins and cash flow.12 Accelevation's pricing fits that prediction. The demand was real, but margin and cash-flow quality decided the price.

Accelevation's own comparison stock also helped explain why it tried for a high price. Forgent Power Solutions, which makes electrical equipment for data centers and power grids, had gained about 40% since its February IPO.12 That gain suggested the market would pay up for this category. Accelevation's result shows the gap between pricing a company in February and pricing one in late September.

Life after the listing

The first month of trading was weak. Bloomberg's figures valued the company at about $4 billion at the $18 IPO price.6 The stock then slid. Around October 20 it traded at about $12.20 and reached a low of $11.58.22 Over the following week it rose about 27% and was trading around $15.55 on October 29.21 On one day that week it gained 11.9%.28

That late-October rebound followed the end of the research blackout on the underwriters. On October 26, Goldman Sachs started coverage with a buy rating and a $22 target. Morgan Stanley started at equal weight with an $18 target, and Baird rated the stock outperform with a $16 target.30 Bank of America and Barclays both set $20 targets.28 The consensus target of about $19.43 is only slightly above the IPO price.30 For a deal these banks ran, that is a modest endorsement.

Reported market caps vary widely after the debut. Some data providers list about $1.74 billion to $1.8 billion based on roughly 120 million Class A shares.2123 Another lists about $2.79 billion.22 Reuters' $3.92 billion figure at the open reflects a broader count of shares.8 Investors should treat quoted valuations carefully until the company's share structure is clear.

The balance sheet changed quickly after the listing. On October 13, an Accelevation subsidiary signed a new $650 million senior secured credit facility with Bank of America as administrative agent. The facility includes a $250 million term loan and a $400 million revolving credit line.25 The proceeds, plus cash on hand, paid off the previous credit agreement in full.21 Insiders also bought shares at the offering price. CEO Michael Rubiera disclosed about $3.7 million of purchases.21

What it means for the IPO pipeline

The coverage agrees on the main point. Accelevation's $18 pricing reset the benchmark for AI infrastructure companies planning to list. Analysts said other candidates may have to accept lower prices, sell fewer shares or wait.2 Some commentators saw it as a sign that private equity owners holding similar assets should lower their exit expectations.3

The coverage disagrees on how bad that is. Some writers called the deal a success that happened to come with a discount: a profitable, fast-growing company still raised half a billion dollars.1 Others saw it mainly as a cash-out for the private equity owner.20 Both views have support. In my judgment, the more useful conclusion is that the market separated the AI buildout from the price of owning a piece of it. Investors still believe in the demand. What they rejected was paying a premium multiple for a low-margin, heavily indebted supplier whose IPO mainly funded its owner's exit. Sponsors with similar companies should expect the same treatment, at least until rates come down.