Data center company Switch, previously an ABS issuer, was enjoying a warm welcome on its first ever CMBS transaction in late January when news about DeepSeek's AI developments interrupted proceedings.
Understandable, perhaps, given this sent tech stocks tumbling amid apparent uncertainty over future data center demand.
Yet one investor told GlobalCapital that he dropped from the deal not because of DeepSeek, but because the sponsor would not provide any information on the tenants — a key part of credit analysis for CMBS buyers.
Ultimately, the transaction drew strong demand and the sponsor was able to raise $2.4bn after only a short delay. But with CMBS expected to take a greater role than before in financing the sector, it's time for investors to demand more detailed, standardized rent rolls for deals.
For information security, privacy, and competition concerns, many data center sponsors do not disclose tenants. And rent rolls have varying levels of granularity which makes data center deals overall much more opaque than other property types in a CMBS.
The disconnect even depends on the type of CMBS financing. Data center conduit loans have standard rent roll requirements, while SASB data centers are much more flexible.
But better standards would benefit everyone: they would allow investors to better analyze the property, understand its complexities, and be able to place more confidence in a transaction.
Changing credit story
In a world where the credit story around data centers was nearly impeccable, it made sense that investors were more flexible on structures and what they would accept. The prevailing view was that the demand was so strong for this property type that it would more than make up for any shortcomings such as higher leverage or lower debt coverage ratios.
But DeepSeek took the market by surprise earlier this month, and some cracks formed in the narrative. They may be small cracks, but they're enough of an idiosyncrasy that investors may start to look at it differently and want more information around data center demand.
Irrespective of AI, the sector is poised to continue benefiting from cloud computing and transitioning to a digital infrastructure. However, much of the excitement also comes from large language models (LLMs) and generative AI which, until now, were expected to require massive data centers from large energy requirements.
What DeepSeek did was to call into question growth assumptions around data centers. If less resource-intensive AI is possible, participants need to recalibrate their expectations and ask for the right information to determine if a particular data center is well-positioned.
Scale back
For now, many view data centers as the most favored commercial real estate sector. If one tenant’s lease was to roll over, many think it would be quickly replaced by another tenant.
Yet if AI-related companies scale back their data center needs or adopt a 'do less with more' mentality, it will affect outstanding and future data centers.
The impact is not only limited to just generative AI and training models. It also intersects with the broader trend of companies going digital. Many companies are deciding to integrate AI into their platforms — the risk of more efficient AI would mean they would need to lease less space in the future which would impact data center demand.
Investors can’t precisely account for these factors unless rent rolls are more granular and standardized.
Only labeling a tenant as “investment-grade” — with a lease type, duration, and power requirements — fails to provide business information for investors to determine if a tenant is at risk of downsizing their leasing space in the future.
While there is a balance between rent roll transparency and the sponsor's and tenant's desires to maintain privacy, information security and competitiveness, investors should stand up and demand standardized rent roll disclosures across all data center deal structures.
These properties can be securitized as an ABS, or as a SASB or conduit CMBS. What is disclosed within each route is a different patchwork of information.
Investors wanting to see more detailed rent roll information is a natural byproduct of a more evolved asset class. Investor confidence and the market overall would benefit from standardization.