Deal Teardown: New Relic, and the Cloud Margin Behind a $6.7B Take-Private
This analysis uses only public SEC filings: New Relic's annual reports for fiscal 2020 through 2023, its quarterly report for the period ended September 30, 2023, and the definitive merger proxy filed September 25, 2023. I have no affiliation with New Relic, Francisco Partners or TPG, and no non-public information about the company or the transaction. Where I estimate, I say so.
New Relic is the most instructive public take-private I know of for anyone who prices cloud cost in a deal. In the three years before it was sold, the company moved its platform out of third-party data centers and into public cloud. Its gross margin fell more than fifteen points and then mostly recovered, while it signed half a billion dollars of commitments that outlived the move. The buyers then underwrote a plan that treated the margin problem as solved.
None of that was hidden. It is all in the filings. Here is what an outside-in cloud diligence pass finds, and the questions it leaves for the data room.
The deal
Francisco Partners and TPG signed a merger agreement on July 30, 2023, at $87.00 per share in cash, and the deal closed on November 8, 2023. The proxy puts the total funding required at about $6.67 billion, backed by a $4.02 billion equity commitment and up to $2.65 billion of committed debt.
The price drifted down during the process. The sponsors proposed $95 per share in February 2023 and $90 in June, and submitted the final $87 bid on July 11. The proxy attributes the decline to revenue rather than cost: consumption growth was decelerating, and by July management expected fiscal 2024 revenue growth near 8% against a street consensus of 11%.
That is the headline story. The cloud story sits underneath it, in the gross margin line.
What the filings show
New Relic's fiscal year ends March 31. Dollar figures are in millions.
| Fiscal year | Revenue | Cost of revenue | GAAP gross margin | Capex | Purchase commitments |
|---|---|---|---|---|---|
| FY2019 | $479.2 | $77.4 | 83.8% | $43.3 | $33.5 |
| FY2020 | $599.5 | $103.2 | 82.8% | $58.2 | $64.7 |
| FY2021 | $667.6 | $181.6 | 72.8% | $18.7 | $494.6 |
| FY2022 | $785.5 | $256.3 | 67.4% | $5.8 | $549.4 |
| FY2023 | $925.6 | $246.4 | 73.4% | $3.1 | $518.3 |
Capex is cash spent on property and equipment. Purchase commitments are non-cancelable future payments at fiscal year end, which the company describes as primarily related to hosting services and software subscriptions.
The 10-Ks tell the story in their own words. In fiscal 2021 the company said it intended to migrate its entire platform from third-party data centers to public cloud providers, and warned that this would mean significantly higher costs in the short term while it paid for both the migration and the data centers it had not yet left. Hosting-related costs rose $70.1 million that year and another $75.9 million in fiscal 2022. In fiscal 2023, cost of revenue finally fell, by $9.9 million, which the company attributed partly to leaving data centers and partly to credits received from its public cloud providers.
By the quarter ended September 30, 2023, just before close, GAAP gross margin was back to 78.3%, with the migration described as mostly complete.
Finding 1: The migration cost about $275 million of cost of revenue
In fiscal 2020, cost of revenue was 17.2% of revenue. Had it stayed at that ratio, New Relic would have spent about $275 million less on cost of revenue across fiscal 2021 through 2023: roughly $67 million, $121 million and $87 million in each year.
Some of that was a trade rather than a loss. Capital spending fell from $58.2 million in fiscal 2020 to $3.1 million in fiscal 2023, about $147 million less over the three years than if it had stayed at the fiscal 2020 level. Servers the company no longer bought became cloud bills it now paid.
But the trade did not net to zero, and where the cost landed matters more than its size. Capex sits below gross profit. Cloud bills sit inside cost of revenue. Moving from one to the other pushes cost directly into gross margin, the metric software companies are valued on, during exactly the years the business was being marketed. Anyone benchmarking New Relic's fiscal 2022 gross margin of 67.4% against observability peers was looking at a migration, not a steady state.
The diligence point: a gross margin in the middle of a cloud migration is not the business's margin. You have to rebuild the run-rate margin from cost of revenue that excludes the overlap: the data centers still running, the duplicated environments, and the one-time migration labor. The filings let you see that the overlap existed. Only the data room tells you how big it was.
Finding 2: The plan assumes the margin is solved
The proxy discloses two versions of management's long-term model: the initial model shared with the sponsors in April 2023, and an updated model the board approved in July and did not share with them. In both, gross margin is close to 80% in every year from fiscal 2024 to fiscal 2028.
In the updated model, revenue grows from $1,003 million in fiscal 2024 to $2,282 million in fiscal 2028. Cost of revenue grows from roughly $202 million to $457 million. Revenue more than doubles, and the margin does not move.
The proxy notes the projections use non-GAAP measures, so the fair comparison is New Relic's reported non-GAAP gross margin. That was 75.9% for fiscal 2023 and 79.0% for the fourth quarter. So the plan starts from the exit rate of the most recent quarter, adds about a point, and holds it there.
That might be right. But a flat 80% across 2.3 times the revenue is a strong claim for this business, for two reasons the filings make plain.
First, New Relic prices on consumption, measured by provisioned users and by the amount of data customers send it, and in fiscal 2022 it said it believed it offered the lowest data ingest prices in the industry. On that model, revenue and hosting cost both scale with data. Margin holds only if the cost of storing and querying a unit of data falls at least as fast as the price charged for it. Scale does not hold margin automatically the way it does for seat-priced software.
Second, the company's 10-Q notes that its cloud costs depend on prices negotiated with providers, "which in certain cases are also our competitors." Every major cloud provider sells its own monitoring service. Your supplier also competing with you does not break the model, but it is a counterparty risk to price in.
The diligence point: the cloud assumption that carries the most value is not a cost line. It is the gross margin held flat across a five-year plan. Test it bottom-up, as cost per unit of data ingested over time against price per unit of data ingested, not top-down as a percentage of revenue.
Finding 3: $518 million of commitments that outlast the migration
Purchase commitments jumped from $64.7 million at the end of fiscal 2020 to $494.6 million a year later, the same year the migration began. At March 31, 2023, they stood at $518.3 million, about 2.1 times that year's cost of revenue. The schedule:
- $88.3 million due within one year
- $276.8 million due in one to three years
- $153.2 million due in three to five years
By September 30, 2023, the total was $463.8 million.
The middle tranche averages about $138 million a year. For comparison, total cost of revenue, including support staff and everything else besides hosting, was running at about $211 million a year in the quarter before close. The commitments also include software subscriptions, so not all of this is cloud. Even so, a large share of the hosting cost base was contractually fixed for years, and it was fixed before revenue growth slowed to 8%.
This is the most common cloud finding in software deals. A company commits to its provider at the growth rate it expected when it signed, which was usually higher than the growth rate it actually achieved. If usage runs below the commitment, the shortfall is paid with nothing to show for it.
The diligence point: ask what usage forecast each commitment was sized against, model usage under the buyer's case rather than management's, and treat any expected shortfall as a debt-like item in the purchase price. Then check whether the commitments can be re-tranched or moved between services. Providers often have more flexibility at a change of control than the contract suggests.
Finding 4: Credits in the recovery
The fiscal 2023 decline in cost of revenue was attributed partly to credits received from public cloud providers. Migration credits are common and legitimate, but they are one-time by nature. If any of the margin recovery in the year before the sale came from credits, the run-rate margin is lower than the reported margin by that amount.
The diligence point: get the credit schedule by quarter, and rebuild gross margin without it.
What a point of gross margin was worth here
In its fairness analysis, the board's financial advisor selected a range of 20 to 30 times calendar 2024 EBITDA, based on comparable public companies. On fiscal 2024 plan revenue of about $1 billion, one point of gross margin is about $10 million a year of gross profit, which is $200 million to $300 million of value at those multiples. By fiscal 2028 it is about $23 million a year.
That is why the cloud work belongs in the price, not in a post-close optimization deck. A two-point miss on the margin assumption, which is well within the range this company moved through in the three years before the sale, is worth hundreds of millions of dollars of value.
The questions I would have taken into the data room
- What was run-rate gross margin in the last full quarter excluding legacy data center costs, duplicated environments, migration labor and provider credits, and what was each of those, in dollars?
- What is cost of revenue per unit of data ingested, by month, for the last eight quarters, and what is it against the price per unit of data ingested over the same period?
- Which commitments were signed when, against what usage forecast, and what is the shortfall under the buyer's revenue case?
- Which provider holds which commitment, and how much can be re-tranched, reallocated between services or extended at a change of control?
- Which legacy data center contracts and leases remain, when do they end, and what is the exit cost?
- What drives the move from 79% to 80% gross margin in the plan, and what has to be true about unit cost for 80% to hold at more than twice the revenue?
What the outside view cannot tell you
Plenty. The filings do not break hosting out from the rest of cost of revenue, do not split spending by provider, and do not show utilization, commitment coverage or unit cost. Purchase commitments include software as well as hosting. I have no view of how any of the questions above were answered in the actual process, or of the company's performance since it went private.
None of this says the deal was mispriced. It says where the cloud risk in the price was, and that every piece of it was visible from public documents before anyone opened a data room. A buyer that starts from that position walks into diligence with the right questions already written.
Teardowns like this one use only public information. For a target you are actually evaluating, the same analysis is done from the data room, with the cloud bill itself. Related reading: the Cloud Due Diligence Checklist for the full question set, and Translating Cloud Infrastructure into Valuation Multiples. Book a free discovery call.