Description

IRM SHORT

We recommend shorting IRM.  IRM has been written up on this site several times, and we suggest reading those writeups for background.  

90%+ of the company’s EBITDA is generated from its core document storage business where it stores boxes of paper in its facilities globally and provides related services such as transportation, shredding, digitalization, etc.   The balance sheet is levered 6x a highly adjusted EBITDA number (management claims it is levered 5x).  Presented in this light, we think most people would agree that this business has terminal value questions and a bad balance sheet.  

Importantly, the company has very weak cash flow generation relative to the company’s highly adjusted profitability metrics and large cash needs.  We believe this is underappreciated with the sellside and market blindly using the company’s definition of adjusted EBITDA and adjusted funds from operations (AFFO).  The stock is much more expensive than it appears when parsing through management’s significant and recurring addbacks as well as recurring cash expenses that are run through the company’s cash flow from investing section of its financials but are conveniently excluded from the company’s definition of both adjusted EBITDA and AFFO.   

The dividend, which is currently yielding 3%, is NOT covered according to our calculations, and we think the company is adding to already high debt levels by aggressively spending capital and doing M&A, ostensibly to appease the market and grow EBITDA.  We believe this is a dangerous strategy given the core paper document storage business will almost certainly decline at some point.  Every industry participant we have talked to openly discusses an inevitable decline in document storage volumes and the business being a “melting ice cube” over time.

The market also seems very excited about the company’s Data Center business.  This seems misguided as it is only 10% of EBITDA.  This business primarily grows through incredibly high levels of capital spend.  IRM is spending >20% of total revenues on capex in 2023 and 2024, with most of that going to new data center builds.  The company’s returns on capital here have been poor with a historical ROA of ~5%, much lower than management’s promised returns in the HSDs on an unlevered basis.

Not only is management spending more than they make to invest in the Data Center business, but they have also been acquisitive to diversify out of the core paper storage business.  In early 2022 they acquired a business called ITRenew in the asset lifecycle management space.  That business is more volatile and has much lower margins than the core document storage business.  The business has also missed management’s revenue guidance by an eye-popping 50%+ over 2022 and 2023.  With the company spending capital well in excess of its cash flow generation by our calculation, their capital allocation has been questionable.

Lastly, we believe that core document storage businesses volumes are also worse than the reported figures quoted by management.  Management claims volumes are flat, but in our view they are being obfuscated by constant changes in disclosure and boosted by growth in lower margin and non-core segments.  We believe that core North American and Wester European document storage volumes, that drive the majority of EBITDA of the core document storage operations, are likely negative.  It is our view that the market would value this stock quite differently if it knew that core profit driving volumes were declining, even if offset by price.  Cable companies are a good analog where volume declines have resulted in discounted valuations despite healthy pricing.  

IRM management has recently been taking significant price increases in the document storage business to show growth, but we believe this is unsustainable, and we expect that these price increases will be harder to justify in a lower inflation environment.  Management claims customers take no issue with their significant price hikes, but their NPS score, which they disclose at 14 is low, and down from 16 the prior year.

Despite long-term structural challenges, profitability metrics that we believe overstate economic reality, weak cash flow, and poor capital allocation, the stock is trading at a peak valuation.  We pencil out a real free cash flow yield of <3%, on a highly levered B/S.  We believe the stock offers ample downside from here given the nature of the core business and disconnected valuation.

Cash Generation is Weak

As some of the structural challenges have been discussed in prior writeups and are likely more appreciated and debated, we will start by discussing the weak cash flow generation that we feel gets less airtime.

Cash flow from operations or CFO (that is pre any capex) has hovered ~$1bn over the last 6 years, as shown in the table below.  CFO was $1.1bn in 2023, which is a step up from it being consistently <$1bn since 2018.   Since 2019 CFO is up 15%, sharply lagging the company’s “adjusted” EBITDA being up 35%.

Importantly, this $1.1bn in 2023 reported CFO does not factor in recurring cash expenses, that show up in the cash flow from investing section and are not capital expenditure related.  Specifically, these expenses are labeled “acquisition of customer relationships,” “customer inducements,” and “contract costs.”  The company lumps all three costs together in its segment financials.   In general, these are cash costs associated with attaining new customers and/or buying out new customer’s existing contracts from competitors.

These costs show up in the cash flow from investing section of the cash flow statement and are amortized on the PNL.  The issue is that the company adds back the related amortization in its definition of “adjusted” EBITDA, therefore overstating EBITDA.   Software companies generally amortize sales commissions over 3-5 years, but the actual cash costs are included in cash flow from operations and calculations of free cash flow.  For IRM, these cash costs show up in cash flow from investing section separately from capex and are therefore ignored.   

As shown in the table above, these “cash flow from investing” expenses to acquire new clients have run between $75-$125mm per year over the last several years and were ~$100mm in 2023.  Therefore, if we take 2023 CFO of $1.1bn and adjust for these costs, you are left with $1bn in underlying cash flow before any spend in capex.  Note that 2023 was the best cash flow year of the last five, and for prior years underlying cash flows from operations were $900mm or lower.

To put this in context, the ~$1bn of underlying CFO pre-capex in 2023 compares to a $23bn market cap today.  That’s a whopping 4% yield before even $1 is spent on capex.   This is for a company that stores paper and has to spend to substantial capital that it needs to borrow to grow its data center business.  You then have $12bn of net debt sitting on top of that $23bn.  2023 interest expense was ~$600mm, so unlevered you are talking ~$1.6bn of cash generation vs. that $12bn of debt (again, pre-capex).   We think true leverage is much higher than the management team portrays.  Management says their leverage is 5x on a lease adjusted basis.  But simply taking the $12bn of net debt over management’s 2024 EBITDA guidance of ~$2bn, shows that even on their heavily adjusted EBITDA numbers leverage is 6x.  In reality, it is even higher as adjusted EBITDA is overstated.

As a sidenote, no sellside analyst ever discusses real cash flow for this business, instead relying on a what we believe is an inflated AFFO metric.  The company puts out 100 pages of material every quarter (press release, detailed supplemental info packet, a presentation) but you have to wait several days for the 10-Q/K to see a real cash flow statement.  They publish an AFFO metric as a proxy for free cash flow, but that adds back numerous real costs and includes a maintenance/recurring capex estimate that seems far too low to us.  More recently, AFFO has benefitted from increasing stock-based comp (added back) in lieu of cash comp, and flat to down maintenance capex, despite the business growing and capital spending increasing rapidly.  Management also recently made the decision to now add back the amortization of the “contract costs” to further flatter their AFFO metric.  

Business is Highly Capital Intensive;  Dividend is Not Covered and Company is Burning Cash

To reiterate, underlying cash flow pre-capex was ~$1bn in 2023 taking reported CFO and subtracting out recurring customer acquisition, inducement, and customer fulfillment costs that run through the cash flow from investing section.

For context, the company plans to spend $1.5bn in capex in 2024 and spent $1.3bn in 2023.   That’s a -$500mm cash flow hole ($1bn in underlying cash flow less $1.5bn of capex).  The company pays a $750mm dividend.   So all in, that’s -$1.3bn in cash burn.   From our vantage point the dividend is nowhere near covered and debt levels must continue to rise to fund management’s plans.  

The company skirts its cash flow scrutiny through aggressive addbacks, but also by claiming maintenance or recurring capex is only ~$140mm.  Recurring or maintenance capex is a calculation that is included in their definition of AFFO.  They claim their recurring capex number hasn’t budged over the last 6 years despite spending $4bn in aggregate on capex and $1bn in M&A over that time.  It’s hard to see why maintenance capex would be flat with PP&E going from $4.6bn to $6.3bn from 2018-2023.

As a sidenote, the company also spends on investments in JVs related to its data centers, as shown in the table above.  This is another $15-$75mm of spend on top of the defined capex budget (i.e., in addition to the $1.3bn in capex in 2023).

In 2020, which is one measure for real maintenance capex, the company spent $195mm of capex in its document storage operations alone.  That would take FCF to $800mm by our math.  This assumes capex in their data center and corporate segments is zero which is unlikely.  That results in a 3% yield on equity with $12bn sitting debt above that.  And those debt levels are only going higher as the company is spending more than it makes.  

That said, it is unlikely that the $195mm of document storage capex spend in 2020 is a sustainable number for the whole company.   Looking historically, the company has persistently spent $300-450mm+ in total capex.  Even going back to 2013, before it acquired its largest competitor, and started spending material amounts of capital on its data center business, IRM spent nearly $300mm+ in capex every year, as shown below.  

This can also be seen if we just look capex excluding all data center segment capex, assuming 100% of data center capex is for growth, which it is not.  Again, the core document storage business alone has had capex averaging $300mm+ and close to $400mm last year.

Putting it all together, we believe real recurring capex excluding the data center growth is probably $300-400mm.  The document storage business has consistently spent this amount alone, despite consolidating facilities, and showing flat volumes according to management (i.e., no growth).  That means real underlying FCF here is likely $700mm or less ($1bn of underlying CFO less $300mm+ of recurring capex), which in turn implies the $750mm 3% yielding dividend isn’t even covered on an ongoing basis.  This is contrary to management’s assertion that the dividend payout ratio is in the low to mid-60s, as they use an AFFO metric that adds back a lot of recurring costs and uses a maintenance capex number that is far from realistic in our view.

Adjusted EBITDA is not Reflective of Economic Reality

The sellside mainly follows the company’s “adjusted” EBITDA calculation for valuation purposes.  On “adjusted” EBITDA guidance for 2024, the stock is trading ~15-16x.  This is well above historical multiples of ~12x.

We observe several flaws with the company’s definition of “adjusted” EBITDA.  

The first is that it excludes the $100mm or so of “customer acquisition, inducement, and contract costs” discussed above.  These should be deducted from EBITDA as they are real expenses.

The second is that there are constant restructuring and integration costs.  As shown below, these costs totaled $200mm in 2023 and have averaged ~$150mm per year over the last five years.  The company has guided to adding back a similar level of 2023 costs to 2024.  We suspect the company is being very aggressive in adding these back costs.    

If we deduct both the recurring restructuring costs + cash costs of acquiring new customers (i.e., contract costs, inducements, etc.), EBITDA is 15%+ lower.  That is very material in the context of the company’s balance sheet leverage.  Because of the leverage, a 15% reduction in EBITDA translates into a 25% hit to the equity value.  One could argue the EBITDA multiple should also be lower given the quality of earnings implied by that EBITDA has been lower.

Digging into the restructuring and integration costs further, one can see above that the company has had varying degrees of addbacks for a decade but that these addbacks have also been trending higher under the new CFO (started in 2020), who we view as aggressive.  The new CFO, the company has been increasingly reducing disclosures and increasing the extent and amount of EBITDA adjustments.

In addition, the company ended a large multi-year restructuring program called Project Summit in 2021.  The aim of Project Summit (10-K description below) was to simplify their global structure and modernize their service delivery.  Project Summit, resulted in restructuring costs addbacks of $450mm in total spread across 2019-2021, and the company estimates it generated $375mm in cost savings.  

That said, the company’s disclosure shown below for Project Summit shows that 90% of the restructuring expenses being added back were for Professional Fees and Other.

Conveniently, as soon as Project Summit ended, management introduced Project Matterhorn.  Management describes this program as more of a change in go-to-market, or a salesforce re-org (see discussion from 10-K below), not a typical cost out restructuring where 1x expenses get added back.  As such, there are no direct cost savings to be had from the project, unlike Project Summit.  

When management introduced Project Matterhorn at its 2022 Investor Day, they casually said that the project will cost ~$150mm/year for four years, or $600mm in total.  This $150mm/year is now blindly being added back by the sellside.  In 2023, management added back ~$175mm of restructuring costs and $25mm of integration costs, so ~$200mm in total.   They are projecting something similar for 2024.  

Management has stated that a large portion of the addback is for reductions in force.   If that is true it makes little sense why the company would spread these layoffs over three to four years, harming the morale of the company.  Especially after three years of Project Summit.  We further question why are there no cost savings associated with this project if they are for RIFs?  In addition, the employee count went up from 2022 to 2023 in the first year of Matterhorn with $175mm of restructuring addbacks.    

As shown below, 10-K disclosure shows that the vast majority of Matterhorn restructuring costs added back in 2023 fell under “other transformation” costs, which again represent professional fees and other.

We suspect that a lot of these addbacks in Project Matterhorn are normal course expenditures that management is adding back to inflate EBITDA and AFFO.  We believe, given their recurring nature, and our observation that these addbacks are going up as a % of EBITDA, they should be taken into account when looking at the value and cash generation of the company.

Poor Capital Allocation

While management talks about document storage being a growing business, we believe their actions show that they know it has long-term existential issues.  They are dumping significant amount of capital into: 1) their Data Center business (currently ~10% of EBITDA) and 2) M&A in their asset lifecycle management (ALM) business.  The ALM business is one where the company handles the disposition of IT hardware and equipment on behalf of its customers, much of it on consignment, thereby exposing it to used equipment/component prices.

The issue we see with the Data Center business is that it is highly capital intensive with low returns on capital.  As discussed above, IRM is borrowing to grow it.   Their average cost of debt is 7.1% as of Q4 23.

One way to measure the returns is to look at 2023 Data Center segment EBITDA vs. disclosed Data Center segment assets.  We have subtracted out spend on properties in development from total assets to make the numbers more apples to apples.  You can see that the pre-tax ROA = 5%.

Another way is to add all Data Center capex + M&A since 2015 when the Data Center business was just starting, subtract out the amount spent for construction in development, and compare it to current run rate EBITDA.   That also results in 20x creation multiple or ~5% pre-tax ROA.  Even when looking at incremental ROA, we also see that this has been ~5% the last couple years. 

So while the Data Center segment is growing in the 20% range, it is doing so through huge capital spend (most of the $1.5bn in capex this year is going to data centers) at low returns.  This is a costly way to diversify.

It should also be noted that a lot of the data center build here is for just a handful of hyperscaler customers.  IRM only builds the data center envelope.  They are also competing with very low hyperscaler cost of capital in bidding for these projects and so it makes sense to use that ROAs are capped.

The company has also been acquiring their way into the asset lifecycle management (ALM) business and botched their first big deal in our view.  They agreed to pay $900mm+ in total for a company called ITRenew in early 2022.  They guided ITRenew to $450mm in revenues in 2022 and ended up with only $213mm in revenue, according to the 10-K.  That is a massive 50%+ miss.  The company had an existing ALM business before it acquired ITRenew.  In 2023, the total ALM business (including the legacy assets) achieved $177mm of revenue, suggesting ITRenew took another large leg down in 2023.  It seems very likely to us that ITRenew is losing money.  Note that even in good times, the ALM business has mid-teens EBITDA margins vs. the core document storage business having margins in the mid-40s.  This is de-worsification.

While not overly material to the overall financials, the ITRenew deal is indicative to us of the poor capital allocation choices of the management team.  One would think this management team would be paying down debt, not trying to grow via low return capex and lower quality M&A, which adds to its already hefty debt load.  We view management’s game of trying to grow “adjusted” EBITDA to keep up with higher and higher levels of absolute debt as very aggressive given the long-term outlook of its core document storage business.

Disclosure Red Flags with Core Document Storage Volume

As mentioned above core paper storage = 90%+ of EBITDA.  Storage volumes are key here as they get a fee/month/box of paper they store + they provide services on the boxes stored such as transportation to and from the warehouse, shredding/destruction, and some digitalization services. Therefore, volume is the key KPI, in our view.

The company has reported flat document storage volumes for the last two years, of ~730mm SF and projects the same going forward.  That said, we believe underlying trends are worse than what management reports for several reasons:

1. The company used to disclose volumes by geography, including in its higher profit North American and Western European regions, before the disclosure was pulled at the end of 2018.  As shown below, North American volumes started declining in 2017 and continued to worsen through 2018.  Note that these volume numbers included the benefit of the “acquisition of customer relationships” (basically buying out contracts from competitors) the costs of which show up on the cash flow from investing section of the cash flow statement, which according to management at the time boosted organic consolidated volumes by 60bps, and likely more for North America as most of those customer acquisitions were in North America.  Also note that these volumes trends are all presented on an LTM basis as of the quarter, so trends in the actual quarters likely look worse as they are pulling the LTM numbers lower over time.

Starting in 2019, the company pulled back its disclosure and instead only broke volume down between “developed” and “other international” markets.  We believe they likely did this to hide the North America volume declines, and since Western Europe was holding up better consolidate the two to show a lesser decline.   You can see “developed” volumes below.  Note that these volumes include the benefit of both outright M&A as well as the benefit of “acquisitions of customer relationships” (the numbers written in red at the top of the bars exclude M&A but still include the benefit of “acquisitions of customer relationships” in their definition of new sales).  These growth trends are also presented on an LTM basis.  You can see total developed markets volumes were negative, and more so if “business acquisitions/dispositions” are backed out.

While “other international” volumes were increasing 5-7%, they were only increasing 3-4% if we back out M&A (again shown in red at the top of the bars).  

The issue with “other international” volume growth is that they it is much less meaningful than North American and Western European volumes.  

First, when last broken out in 2018, North America and Europe were 75% of volumes.  “Other international” was only 25%.  Second, developed market unit economics are far higher than other international unit economics.  Specifically, North America was 60% of volumes in 2018, but was 75% of EBITDA.  North America and Europe were 85% of document storage EBITDA.  Notably, EBITDA per cubic foot of “other international” was ~55% worse than that in North America.  In other words, North America volumes drive the majority of the earnings power in the segment.  And we know those volumes were declining when last disclosed.  They pulled the geographic disclosure totally in Q3 19.   We believe management purposefully did not want to disclose volume declines in higher profit regions, knowing that it may weigh on valuation.

Given the negative North America trends in 2018, and negative “developed market” trends through 2019, we have no reason to believe that the paper storage markets suddenly got a lot better for IRM in North America.  

In fact, post COVID paper consumption has been sharply lower, as shown below.  Conversations with industry participants have suggested increased work from home post COVID has resulted in less office paper generation, which will weigh on developed market incoming volumes.  Not to mention things like Docusign, the cloud, etc.  Therefore, we believe that core, higher profit, volumes in North America and Europe are declining.

2. Total volumes disclosed have also been helped by growth in non-core Consumer volumes.  The company used to break down its 730mm cubic feet of volumes by end market, distinguishing among core document storage volume, entertainment/art storage volume, and consumer storage volume.  They pulled this disclosure in early 2022.   

Consumer volumes represent volumes from a JV w/ Makespace/Clutter where IRM provides back-end storage services in its facilities for the JV.  As you can see below, consumer volumes had been growing from Q4 17 through Q1 22 when last disclosed.  If we back out these consumer volumes, as well as the impact from Russia deconsolidation and M&A, we can see that total volumes ex-Consumer were declining ~1% from mid 2020 through Q1 22.  Again, another data point that suggests core storage volumes are likely weaker than what management is showing.

On the Consumer volumes, they disclose the JV revenue generated in their annual filings and you can see the price/cubic foot on these volumes is inferior to document storage.  We also know the JV loses money and Clutter was on the verge of bankruptcy before being bailed out and merged w/ Makespace by IRM.  We therefore believe it is appropriate to look at the volumes ex-Consumer.

Significant Recent Price Increases Not Sustainable

Management projects the document storage business to grow 6% in 2024 after growing HSDs in 2023.  Management breaks down this growth as being driven by “flat volumes” and then M-HSD “revenue management” which is basically price increases.  We would argue that given the volume declines in North America and Europe and mix shifts (Consumer) discussed above, real pricing is likely higher than the implied M-HSD level management discusses.   While we admittedly do not know when the pricing runway ends, we have consistently heard that these increases were easier to take in the more inflationary environment of the last few years.  We also believe that these price increases are falling disproportionately on smaller customers as larger enterprises have more negotiating power and better contracts.  We believe that this of level pricing gains are likely unsustainable over the long-term.

Management is Promotional, But Speaking with Their Wallet

Management is highly promotional.  They talk about their business as a growth business, about the dividend being well covered, about leverage being the lowest in years and within their target range, etc.  That said, the CEO is voting with his wallet and unloading 1mm shares (current share prices = close to $80/share)

Valuation

We can debate the multiple for IRM, but even if this were to trade at a very healthy 20x “real” FCF of $700mm (on 6x+ of leverage no less) the stock would have an equity value of $14bn and have 40% downside.

Catalyst

Volume declines

Weak cash flow