Long-lived parts and maintenance revenue is an underappreciated benefit for many behind-the-meter power firms, some of which have gotten very cheap on 2027-28 estimates. INNIO expects more than $50M a year per GW in services from its 19GW backlog and reserved slots, much of it arriving five and ten years after deployment, atop roughly $30M a year per GW from its 44GW installed base. That revenue could persist long after the data-center capex boom ends. x.com 

Bloom Energy could look cheap if it hits its revenue targets amid exploding demand for on-site power. Revenue is projected at $4.5B in 2026 and $16B by 2029, implying 3.5x 2029 sales at a $56B market cap. It ranks alongside Forgent Power Solutions, perhaps better, as a buy at current levels. x.com  x.com 

Contracted AI compute demand makes four AI infrastructure stocks worth buying at 52-week lows. Applied Digital has a $36B backlog and expects 600+ MW online within 12 months; Cipher Mining has $8.5B+ in AI leases with payments ramping; CleanSpark has a 20-year $6.6B lease and 1.8+ GW power portfolio; CoreWeave has a $104B backlog and raised its year-end ARR target to $18.5-19.5B. x.com  x.com 

INNIO looks attractive despite sponsor-selling pressure: it expects roughly the same 2026 revenue as Bloom Energy at about one-sixth its market cap, while backlog rose from $1.0bn to $6.6bn and customers prepaid $1.4bn. Its IPO was entirely secondary, sponsors still own 86%, and the November 30 lockup expiry could bring another block sale before business growth shows through in the stock. x.com  x.com 

INNIO is an under-the-radar power play that looks cheap next to Bloom Energy. It has a $6.6B backlog and almost 4 GW of annual capacity by 2027, at a $23B market cap versus Bloom's $70B with roughly the same capacity, though the comparison isn't that simple. It isn't on anyone's radar yet, which is a great thing. x.com  x.com 

Bloom Energy's results show hyperscalers paying for faster data-center power sooner than modeled. Revenue hit $1.24B versus $1.05B expected, and 2026 guidance rose to $4.4-4.6B; S&P 500 inclusion brings index buying. Operating cash flow was slightly negative as capacity expands. Bloom is not cheap after a 160% run and expectations are high. Overweight, $334 target with 40% stated upside. x.com  x.com 

Applied Digital still looks cheap on its buildout, though it needs momentum and cooperative bond markets. Its CY26 deliveries may miss expectations by ~50MW, with a catch-up in early 2027; it plans 600MW in service over 12 months. Two expansion leases totaling 250MW are expected by year end at >15% higher pricing, later than first signaled. Macquarie is owed shares and ~25% of NOI on partnered North American deals; Finland's first 100MW is expected in 2028. x.com 

Solaris Energy Infrastructure is a very cheap way to play the US power build-out, now expanding to 3.1GW on booming demand from customers like xAI. Two deals add 900 MW of turbines: 400 MW via the Genco Power acquisition and about 500 MW of turbine delivery slots. The $1.55B cost, about $1.73M per MW, is well above prior purchases, but the company sees clear line of sight to deploy. Stifel estimates the added 900 MW could lift 2028 EBITDA about $300M above the $760M expected, about $15 a share. … Show more

Fluence's batteries are the only grid bridge shipping now, with gas turbines sold out into 2030 and transformers on 130-week lead times. Investors fixated on the August guidance cut are missing about $850M of data-center business booked through July, including a $300M behind-the-meter order. It trades at 0.6x EV/sales on $2.63B of revenue growing 8%, with net debt of $74M against $339M cash. The SEC investigation and pending Abramov class action could make the discount permanent. x.com 

FuelCell Energy could ride AI data-center power demand with high-temperature fuel cells for clean baseload, scaling 12.5MW FuelCell Blocks for hyperscalers. At about a $650 million market cap, it has $170 million of TTM revenue, Q1 up 61%, a $1.17 billion backlog, $312 million of cash and output ramping to 350MW. Risks are negative gross margins, cash burn, dilution and unproven execution; @SueWallSt notes its 1.5GW of data center deals are proposals and backlog shrank. High risk, high reward. … Show more