AWS 벤더 프로세스 직접 관여자 콜. 2026 Q1 적격화 착수, 13~18개월 소요 전망. 기술이 아니라 '계약조건+스케일'이 핵심 병목임.
AWS는 2025년 3분기 FTAI 인지, 2026년 2월 본격 적격화 시작. 신규 공급사라 기존 Mitsubishi/GE 대비 더 오래 걸리며 풀 적격화에 13~18개월 전망. 현 시점 착공 기준 완료는 2027년경
왜 중요 · FTAI의 2~3GW 캐파 가이던스(2027)와 풀 적격화 시점이 정확히 겹침. 즉 실제 매출 인식은 2027 이후로, 시장 기대보다 늦을 수 있음
1)MSA(상업조건/카탈로그) 2)SKU·모듈러 아키텍처/소싱 3)엔지니어 연결·지역(US/EMEA/APAC) 4)RMA·서비스·워런티·TCO 5)실제 1대 테스트. 기술검증은 4~5단계. 컨버전 실현가능성보다 계약조건 합의가 우선
왜 중요 · 시장은 'CFM56 컨버전 기술 입증'을 핵심 관문으로 봤으나, 실제론 상업조건·스케일이 게이트. 리스 가능 자산 특성상 lead time 길어 상업조건 선합의가 합리적이라는 구매자 논리
VP는 2GW(대부분)를 원하나 비현실적, 실무자는 타 하이퍼스케일러 경쟁 감안 ~500MW 예상. 25MW급 엔진 기준 4기=100MW는 '나에겐 너무 적다'. 더불어 3GW 가이던스 자체에 회의적
왜 중요 · FTAI 파워 TAM에 대한 실수요 측 첫 정량 앵커. 또한 '3GW 약속→5GW 주문받고 백오더'로 lead time 우위가 결국 소멸할 것이라는 구매자 시각
직구매 기준 MW당 가격이 레거시 OEM과 동등 수준, 약간 높을 수도. 가격은 아직 협상 중. 리스 시 MW당 더 비싸짐
왜 중요 · FTAI 차별화가 '가격'이 아니라 '속도(적격화 후)'임을 구매자가 명확히 함. 컨버전 저원가 내러티브가 최종 ASP에는 프리미엄 없이만 반영되는 구조
유틸리티 그리드 연결 전 3~5년 브릿지. PSI(Weichai/3MW급)는 백업, Bloom 연료전지도 미국선 백업·브릿지로만. FTAI는 1차전력 격차를 메우는 브릿지
왜 중요 · 용도가 백업이면 직구매(소유), 브릿지면 리스 선호. 이 포지셔닝이 FTAI 매출구조(판매 vs 리스)와 마진 인식 방식을 좌우함
공장은 '사진만' 제공, 실사 미요청 상태. 캐파 데이터에 베이스라인 없음('25% 증가'의 비교기준 불명). 발표 6개월 지났으나 Azure·Meta 동료들도 신규정보 못 받음. 캐파 실증은 Q4'26, 시장 판매 실증은 Q1'27 예상
왜 중요 · 스케일 검증 불가가 적격화 진전의 실질 장애물. 'CFM56 항공엔진 시절과 비교'는 apples-to-apples 아님 → 캐파 가이던스 신뢰도에 직접 의문
FTAI가 엔진 소유 유지하며 PPA로 전력 직판하는 모델. 단 시간이 더 걸려 실제론 직구매가 빠를 것. 별도로 Vertiv/Siemens 등 전기 통합사 경유 리스 모델도 존재
왜 중요 · FTAI 사업모델이 단순 엔진판매를 넘어 에너지서비스/PPA로 확장 가능성 시사. 다만 단기 실현성은 낮아 옵션가치로만 평가
2GW/2027 약속이 풀 적격화 완료 시점과 일치 → 실주문·매출은 2027 이후. 주가의 'AI 옵션가치' 선반영(MS 목표 $220, EV/EBITDA 26x)과 실현 타임라인 간 갭이 큼
컨버전 기술 회의론은 후순위로 밀려 긍정적이나, 상업조건 합의 실패는 '관계 자체를 파괴' 가능. FTAI가 하이퍼스케일러 요구조건(800V, 고압→저압, 배출규제, 대체연료 소싱)에 얼마나 양보하느냐가 진짜 변수
FTAI의 핵심 셀링포인트인 '속도'는 적격화 후에만 유효하고, 다수 하이퍼스케일러가 동시 주문 시 백오더로 lead time 우위 소멸. 'Google/Meta 등이 동시 적격화하면 신규 벤더로 lead time만 재발생'이라는 구매자 통찰
$15~16M/25MW unit 기준 500MW≈20기≈$300M+ 규모. 다만 리스/PPA면 인식방식 상이. 멀티 하이퍼스케일러 합산 시 3GW 흡수 가능하나 각사 회의론 동반
6/9 AI는 유지로 하향, Chairman은 'Power 안다≠납품·SLA, 선반영' 지적. 이번 AWS 콜은 '1단계 MSA, 데이터 부재, 회의적'으로 다운그레이드 논리를 실증 보강함
구매자가 마지막에 강조: 가스가격 상승 시 하이퍼스케일러는 연료전지·배터리 등 대안 탐색. 비용보다 속도가 우선이나 장기적으로 천연가스 터빈 수요 자체가 흔들릴 수 있음 — FTAI 파워 TAM의 거시 변수
팀은 2023년 초 FTAI를 처음 인지한 뒤 항공 MRO/모듈교환(MRE) 애프터마켓 구조적 쇼티지 테마로 추적해왔음. 2024년 CFM56 모듈화·오픈 애프터마켓 전략과 EV/EBITDA 8→16x 리레이팅에 주목하며 비중을 가져갔고, 2025년 Muddy Waters 공매도 리포트 급락과 WLFC·AER 등 동종 비교를 거치며 검증을 강화. 2025년 말~2026년 초 FTAI 파워(CFM56→데이터센터 가스터빈) 출시를 'AI 내러티브 업사이드'로 받아들였으나, 단위마진·캐파 디테일 부족과 ProEnergy 등 선행 상용화 사례, 선반영된 밸류에이션을 근거로 신중론이 부상. 이에 MSFT·AWS 벤더 담당자 콜로 적격화 실체를 직접 검증하는 단계로 이행했고, 6월 들어 AI 레이팅을 비중확대→유지로 한 단계 낮춤.
6/9 AI 레이팅 하향과 6/11 MSFT 콜의 '선반영·납품 미검증' 논리를 AWS 콜이 '1단계 MSA·데이터부재·캐파 회의'로 재확인
2026-01-05 '단위마진·추가캐파 디테일 부족' 아쉬움이 AWS 측 '베이스라인 없는 캐파 데이터·사진만 공개'로 동일하게 드러남
2026-01-05 GPT 추정 IRR 30%+·가스터빈 전환 매력과 달리, AWS는 ASP가 GE/Mitsubishi와 on par(약간 높음)로 보아 가격 프리미엄 없음 시사
FTAI의 '속도(speed to power)' 셀링포인트를 AWS는 '적격화 후에만 유효하며 멀티고객 동시주문 시 lead time 우위 소멸'로 반박
AWS 단독 현실 주문 ~500MW(25%) 정량 추정, 적격화 완료=2027(캐파 가이던스 시점과 일치), 캐파 실증 Q4'26·시장 실증 Q1'27 마일스톤
FTAI 자체설비(Montreal/Miami) PPA 직판=준-전기통합사업자 옵션, 그리고 천연가스 가격이 수요 스위치라는 거시 리스크 포인트
| AWS 인지 | 2025 Q3 |
| 적격화 착수 | 2026 Q1(2월) |
| 풀 적격화 소요 | 13~18개월 |
| 완료 예상 | 2027(캐파 가이던스 시점과 동일) |
| 현 단계 | 1단계 MSA(착수 후 4개월) |
| 캐파 실증 기대 | 2026 Q4 |
| 시장 판매 실증 기대 | 2027 Q1 |
| 1단계 | MSA — 상업조건/카탈로그/커스터마이징 범위 |
| 2단계 | SKU·모듈러 아키텍처/엔진사이즈(5/25/50MW)/소싱 |
| 3단계 | 엔지니어 연결·전력위치(US/EMEA/APAC) |
| 4단계 | RMA·서비스·워런티·TCO (기술리뷰) |
| 5단계 | 실제 1대 구매·리스 테스트 → QVL 등재 |
| 단가 | $15~16M/unit (협상 중) |
| 엔진 사이즈 | 25MW급 (4기=100MW) |
| MW당 ASP | GE/Mitsubishi와 on par~약간 높음 |
| FTAI 캐파 가이던스 | 2~3GW/년 by 2027 |
| AWS 가능 추가캐파 기대 | 2~3GW (구두, 회의적) |
| AWS 이상적 주문 | 2GW(비현실적) |
| AWS 현실 주문 추정 | ~500MW (3GW의 약 25%) |
| 적격화 비용(AWS) | $200K~400K/건 |
| FTAI 용도 | 3~5년 브릿지 전력(midterm) |
| 판매구조 | 직구매(소유) vs 리스(전기통합사 경유) |
| FTAI 자체설비 | Montreal, Miami → PPA 직판 옵션 |
| 경쟁 신규엔트런트 | PSI(Weichai/중국, 3MW급, 백업용) |
| 레거시 OEM | GE, Mitsubishi(직구매 6개월 적격 가능) |
| 연료전지 | Bloom(미국선 백업), 한국·도쿄선 1차전력 |
| 거시 리스크 | 천연가스 가격 상승 시 대안 탐색 |
녹취 원문 전체를 보존하고 STT 오류·고유명사만 교정(점선=교정, 형광=핵심)했습니다.
Um, hello?
Hello. Sorry for, uh, being late there. Had to get back to my office.
Oh, no. I'm, I'm totally fine. Um, thank you so much for taking the time for today's call.
Thank you.
[chuckles] Um, so to briefly introduce about myself, uh, my name is Eujung Lee, you can just call me EJ, um, from Moss Asset Management, and, uh, we are a hedge fund based in Seoul, Korea. And I'm currently, uh, doing a research on FTAI. And, uh, from today's call, I would like to better understand, uh, how hyperscalers like AWS, uh, is currently viewing, uh, FTAI's emerging power generation offering.
Um, since there has been a lot of investor discussion around FTAI's ability to serve AI data center demand, but it remains unclear where the company actually sits, uh, within the customer qualification process. Like what milestones-
Yeah
... remain and how customers are evaluating its execution capabilities beyond, uh... So, uh, that's why I requested today's call.
Sounds great. Yes, I can definitely speak on that. I think that we are, uh, overall pretty excited-
Mm-hmm
... if FTAI, if FTAI could actually do what they say they can do. There's, you know, one of the things they mentioned is they can add another two to three gigawatts of capacity.
Right.
Uh, and as you're very aware, right, we're in a situation right now where my main constraint is power availability.
Mm-hmm.
Um, and there is-- Even for, like, behind the meter, right?
Mm-hmm.
That's one of the things we've been looking into is behind the meter. And even for behind the meter, uh, I, I have really long, uh, uh, lead times for natural gas turbines and things of that nature.
Right.
So if we could get this extra capacity, it would be great.
Mm-hmm.
Uh, however, I will say, like, for us, um, we are working through a qual-- a po-potential qualification, uh, process.
Uh-huh. Uh, thanks. Um, so I'm curious, like when did, uh, when did you start, first started, uh, reviewing this, uh, FTAI's, uh, aeroderivative engine solution? And like-
Yeah
... how long has this, uh, qualification process been going on?
Yeah. So we first started looking at it in Q3 of 2025. Uh, but the qualification process did not begin till Q1 2026.
Oh.
That's typically when we start qualification processes. Now, FTAI is a totally new-
Mm-hmm.
Will be a totally new supplier, so they're gonna... It's gonna take a longer time-
Oh
... uh, to get them qualified. And it-- we're, we're probably looking at somewhere between, uh, uh, 13 to 18 months-
Oh
... for full-
Oh.
Yeah, full qualification. Um, a-and right, like FTAI is a very new company as well-
Right, right
... um, with their TFM65 engine. So it, it's gonna take, uh, a longer time than it would for like, you know, if I was qualifying new things with like Mitsubishi-
Oh
or GE or, or, or the, or the like.
Mm-hmm. Mm-hmm.
Mm-hmm.
I see. Um, but then, uh, as you already mentioned, uh, this, these other large OEMs, like, uh, somewhat, uh, established players, uh, have-
Mm-hmm
... um, for... Their lead time is quite, uh, like f-
Long
... several like years ahead.
Yes.
Whereas, uh, FTAI argues that their competitiveness comes from like the speed, uh, speed to market.
Correct.
So-
No, but, but it's speed to market once they're qualified, right?
Ah.
Um, and also, uh, one of the things that we, that, um... One of the things that we are concerned about, right, is that we're not the only hyperscaler they can potentially get to, right? They, they can get to other hyperscalers, and they also have, uh, a, a separate business, right? So FTAI is just not for-
Right
data centers, right?
Mm-hmm.
They're for aviation-
Mm-hmm
... and railroad, all those recipes hubs. Uh, so like we wonder would they put data centers as a priority?
Mm.
Uh, and the other thing would be, you know, if, if we get into this market and we qualify them and... But let's say they're also qualified with Google or Meta, would those lead times just now resurface, but now just with this new vendor, right?
Mm.
So I think that right now, yes, that's what their competitive advantage is.
Mm-hmm.
But that could, uh, in, in my eyes, that could easily change.
Mm-hmm. Mm-hmm. I see. Uh, thank you.
So, um, since I'm not an expert in this like, uh, AI data center or this, uh, uh, vendor management, uh, field, um-
Mm-hmm
... so, uh, for a new player like FTAI, um-
Yeah
... is it normal, like is it like an industry standard to go through this 18 months of qualification like process?
Well, it's like usually around, um... It, it could be shorter.
Oh.
But there... I, I should step back and say there's a two-pronged approach.
Mm-hmm.
So we could potentially not work directly with them, right? So we can work through, with them through an electrical integrator. So we can work with like a, uh, a, a Vertiv or Siemens, and then Siemens or Vertiv are actually making the purchase, and then we're leasing the engines through a Siemens, a Siemens or a Vertiv.
So that's one model, and that's like usually the model that I like because I'm just leasing. I'm not actually owning this big CapEx, right?
Mm.
Because these are big, you know, uh, gas turbines or-
Mm
... uh, big CapEx, and we usually would rather lease them. And, and, and that scenario, right, that means like the speed would de-be dependent on how fast they could get qualified-
Right
with these electrical integrators. Um, however, we're stressing a constrained environment that like, hey, I need speed to power. I don't wanna maybe wait on the lease. I wanna buy direct. And if we're buying direct from the OEM, it, it is like, you know, you know, the fastest way that'd be potentially qualified would be like six months.
Oh.
But, uh, that would be with like a, a, a pretty established company. So like, uh, a good example would be Mitsubishi. Mitsubishi really also was not really in the data center market, but we were able to get them qualified fairly quickly, especially in our APAC region, to help us with, uh, gain power in our APAC region. So it, it, it's, it's not impossible to qualify in under 12 months.
It's just that typically that's the timeframe.
Mm, I see. Ah. So it seems that, uh, between those two options, uh, currently FTA is going through, uh, which model? Like which option?
I, I think they're gonna mostly go through the direct OEM model.
Oh.
Right? They have to establish relationships with electrical integrators.
Mm-hmm.
They will have to, uh, kind of, uh, confer to a lot of the, the, uh, things that the electrical integrators want to do.
Mm-hmm.
And I don't think that they will do that quite quickly. So like I think that they will go to that first model eventually, but at least for like their initial rollout of engines to, to, um, uh, you know, not just hyperscalers, but also like co-locations.
Right.
I think that they're going to do a lot of, uh, a lot of direct, you know, like direct OEM-
Mm-hmm
... where we, uh, buy these turbines.
Directly.
But like buying them from a behind the market... excuse me, behind the meter standpoint.
Mm-hmm.
Where we know that like eventually we still wanna get on the grid, but it's gonna take, you know, years to get on the grid.
Right.
Um, then, um, as you mentioned, you first, uh- Became aware, uh, of FTA starting, uh, second quarter of last year. And you-
So more like Q3, Q3 this year.
Oh, Q3. Um, so like second half of last year. And-
Correct
... the actual, uh, qualification process started in first quarter of this year. Like, um-
Yes
... uh, what, um, what, like, what... Is there, like, a moment that, um, made A- AWS, uh, aware, uh, aware of the, um... Like, is there a-a-
Oh, like-
... aha moment that AWS, uh, act- uh, started to, like, review FTA more, like, seriously? Like-
Yes.
Mm-hmm.
Um, I think that, um, we-- So it, it, it's public knowledge, right?
Mm-hmm.
Public available information that last year, uh, well, uh, late last year, our CEO mentioned that we're gonna make huge CapEx investments into AI.
Uh-huh.
So I think that, like, once we had the increased, uh, budget for AI-
Uh-huh. Mm-hmm
... we knew that we could look at... We, we, we could have, like, a, an approach of don't leave any stone unturned, right?
Mm-hmm.
So we wanted to look at any behind-the-meter approach. 'Cause again, like previously, for behind-the-meter generation, we don't like to go, we don't like to work direct. But we saw that... And I think there's also, like, the market is shifting, right? We saw that Meta was doing that. So we said, "Hey, we have to do it too because otherwise we may be left behind-
Mm-hmm
... if we're not developing these, uh, relationships directly with OEMs."
Ah.
And we saw that FTAI, uh, with their big pitch being they can, they can bring speed to power.
Mm-hmm.
So that, you know, of course, that really intrigued us. Uh, so, you know, we, we, uh, start talking, uh, in Q4 2025, and then, uh, that fits through qualification process with our engineers and, you know, like, of course, like, drawing up the contract, legal things, all those things were really happening. Um, that's when we start to really get into it. But I think the real big driver for it is-
Oh
... that we as a company want to get more power.
Oh.
And we are now looking at more unorthodox ways of getting that power.
Oh, that makes sense. 'Cause I was curious whether, um, as you are probably aware of, FTA, uh, announced that they are gonna enter this, uh, aeroderivative power engine business, like, late last year. And it seems that-
Yeah
... they hired some ex-Blue Origin or AWS or Siemens experts into their power business.
Yeah.
So, um, I was curious whether FTA's, uh, stance has been more aggressive in terms of approaching this hyperscaler customers and, like, appealing their-
Yeah
... um, like, capability in terms of the, this aeroderivative engine. Mm-hmm.
And, and, and I think not just them, but, um, uh, a lot of, like, power players in general have become more, uh, aggressive in the, um... Have become more aggressive in the data center market because the data center market is a market that's really exploding.
Right.
Um, so, like, if you aren't investing into that market, I think a lot of the-
[laughs]
... a lot of engine makers and power generators-
Mm-hmm
... uh, you know, you're leaving a lot of... You're, you're losing a lot of potential demand-
Mm-hmm
... um, uh, if you are, you know, in the more older businesses.
Mm-hmm.
So I think, um, one of the main things here was that, that, uh, FTAI team, like, saw that, and then they decided to kind of invest into it. Uh, 'cause it's, of course, what you mentioned. There, there was some R&D investment that had to happen. Um, and, you know, uh, working with people that know about the data center market, know the type of power we want to generate. You know, like, a good example we use is we're moving to eight hundred volt power, and we need, like, data generation, uh, switches and modules, module architecture that can go from high voltage to low voltage.
Mm-hmm.
So those types of things, the things that, like, when we're doing this qualification process-
Mm-hmm
... that's what we want to see.
Mm-hmm.
And I think that, like, I, I hope to see that, right?
Mm-hmm.
And, like, that's my expectation. But, you know, uh, all those things are predicated on them doing their initial R&D research.
Mm. I see.
Then, um, regarding this qualification, uh, like, process and stages, um, are there, like, somewhat, uh, established steps and criteria that AWS, uh, wants to see in order to-
Correct
... uh, and what kind of steps does, uh, A- FTAI has to, like, prove? And currently, at which stage does, uh, FTA, uh, currently, uh, stays at?
Yeah. So right now they're at the very beginning.
Mm.
Um, and that's, like, drawing up a MSA, so, like, a master service agreement about, like, if we were-- Like, we're not guaranteeing any business, but if we were to do business with you, what would be the pricings? What would you be able to offer?
Mm.
So those types of things, they have to be established in the MSA. Uh, because, like, you know, before we start bringing in like testing and all that.
We want to establish an MSA. So that's the first step, and that's where we are right now is establishing MSA, understanding like what is their catalog, what type of customization they will have available and things like that. So we can understand like what can we get from you. So that's basically what MSA is. Like what can we get from you? What type of agreement do you want us to sign?
Like later on, right, so this MSA is just a high level agreement.
Oh.
But like later on, right, we may have to sign a longer term agreement when it comes to like, you know, if we want to do a lease or do we want to do our own. Like how do we want to, you know, do this? So those first, that's like the first step.
Oh.
The second step is with-- the second step would be like for the SKUs that we have agreed on, what type of customization or modular architecture would it be?
Mm-hmm.
So we'll get an understanding, okay, do you have twenty-five megawatt engines? Do you have fifty megawatt engines? Are they smaller five megawatt engines? Like what size of the engines you have.
Do-- are they able to, uh, you know, do a high voltage to low voltage? What are the enclosures like? Where are you sourcing your material?
Mm-hmm.
Like is it sourced from, um, another country? Is it sourced here? Um, we need to know all those things. How will you supply us? Will you have a relationship on the ground for installation? So that's like the second step.
Mm-hmm.
Um, then the third step is working with our engineers to understand the actual connectivity and where we need the power. So like will it be here in the US? Will it be in Armenia-
Oh
... Europe and Middle East? Will it be in APAC? You know, can you serve those geographies? Uh, and then lastly, we'll have an actual contract-
Oh
So we can determine like RMA if, you know, if applicable, right? If we actually do a buy route versus a lease route, we'll do an RMA. We understand like, okay, like can we do customization? Especially if we like buy it, right? We wanna usually have a full customization.
Oh.
So like we have to go through all those things.
Uh, and then like also like at that time we'll have steps on like what is the service agreement, what's the warranty, what's the actual full total cost of ownership. All of that would be established in that fourth step.
Mm-hmm.
And then in the fifth step we would actually like do some testing in which we would buy an actual or lease an actual engine to start to test it in a small amount.
If it all works out, um, then they will be qualified and then we can start having them on our qualified vendor list, then we can start sourcing from them. But it is that five-step process. And that's why it typically can take some time to get a vendor fully qualified and can take up to a year and a half, um-
Mm
... uh, to go through that full process.
Mm. Oh, I see.
So it seems that, um, like the technical review happens at the very last stage.
Well-
Or-
... more like stage four and five-
Four and five
... is like the technical review.
Oh.
Yes.
Mm-hmm.
'Cause the thing is, is like, like during a technical review, um, we can determine still like, okay, like they failed it, but, um, they can still potentially be a qualified vendor or supplier with a different type of SKU, right?
Mm-hmm.
Maybe they just need to change their SKU up or fix it, right? So it won't totally remove them from the running. While like these earlier things like the type of contract and agreements that we wanna have, um, that could potentially like, uh, you know, just totally destroy the potential relationship altogether.
Oh.
So we want to do that early first-
Oh
before we start to work together with engineering and things like that. So-
Oh
... we first have to establish that relationship.
Oh. Because I thought the other way around, like, uh, if and only if like FTA, uh, clears this technical review process, meaning that, um, if only if FTA can successfully show that they have the ability to successfully convert the, uh, retired engine into the data center power solution, then they can move on to this like commercial review and, uh, like, uh, concept.
Well, typically for engines, that would be the case if we were talking about like components for like racks or things like that.
Oh, okay.
But for engines it's a little bit different, right?
Oh.
Because we can lease, right? So like the commercial terms for us are even more important in the very beginning.
Oh.
Because of how long the lead times can be to get-
Mm-hmm
these, to get these engines. It wouldn't make sense for us to kind of work through all that technical work-
Mm
... uh, which can take up to, you know, that can take several months if at the end of the day we don't end up agreeing on commercial terms, right?
Mm-hmm.
It wouldn't behoove us to do that.
Um, but the other way around makes more sense because of the nature of these gas engine turbines, right? Like, um, you know, like for us, like another important thing is the emissions, right? Certain states have emissions laws or like, uh, you know, um- Is there, if, let's say like these engines don't run on natural gas and they run on something else, like on alternative fuel, uh, can we get that alternative fuel?
Will they help us source it? Like some, these terms really matter a lot to us.
Oh.
So we wanna ensure that they can, like, here are our terms. Well, do you even, like do you even agree to these terms that we have?
Mm.
And if so, okay, now we can start to work on the technical stuff.
Oh, I see.
Then, um, would it be safe to assume that, um, it's not much of a matter of whether FTAI can successfully, um, execute its conversion, uh, like converting this engine, but rather how much, uh, like FTA is willing to like cooperate, uh, and meet with this AWS needs and-
Correct.
Oh.
Like that and scalability-
Scalability
... are two things that we are the most concerned about with FTAI, right? Um, would they be able to conform to the terms that we want, especially if it's an ownership relationship where we're like buying an entire engine. And then secondly, would they be able to scale, right? 'Cause I might want a bunch, right?
Mm-hmm.
Can they actually make to our volume that we would, that we would want?
Mm-hmm.
Or are they say, "Hey, you know, we, we..." Like, the volume that they can do is so small that like, oh, it doesn't even make sense for this technical process because they can only get us like maybe 100 megawatts of energy. Well, m- uh, which, which is good—
Okay.
... but like for me, that's not a lot. Like that might be great for like a co-location.
Mm-hmm.
Um, but that wouldn't be enough for me to, to, to maybe spend, right? Like for you, for us, right, we're spending resources and money and, and, uh, full, full-time employee hours to do this, uh, qualification process.
Oh.
So this qualification like, you know, I mean, typically for us, qualification can cost us up to like us up to $200,000 to $400,000. So—
Oh.
... w- with 100 megawatts of energy, it's worth $200,000 to $400,000.
Right.
But like for us, the time also—
Mm.
The time investment may not make sense, right?
Mm-hmm.
So like if, if we, if we see something like that, we might tell them like, "Oh, okay, we'll, you know, we, we won't work through qualifying you. We'll work through a, you know, we can work through a, either a, uh, electrical integrator to electrical integrator qualify you and to do the work. And you're only giving us a little bit of energy anyway, so like it doesn't make sense for us to work with a direct relationship together."
Oh.
Or maybe even, you know, like a distributor relationship that if we see like they can't scale, right, it, it doesn't make sense for us to have a direct relationship.
Mm-hmm. Mm. And I know it's too early, uh— uh, currently, but, uh, since, uh, you have been, since, uh, you mentioned that this MSA process, uh, already started.
Mm-hmm.
Uh, based on your observation and the current discussions y- uh, you're having with FTAI, how would you think of— their, um, like ability and willingness to, uh, meeting the, uh, meeting AWS requirements as well as, um, keeping up with the scalability that, uh, hyperscalers want?
Mm-hmm. Yeah, so right now at a very high, and this is of course, like we mentioned, very early— very high level right now. But we do believe that they can fit our requirements— uh, at a very high level.
Right? So a lot of the contract terms that we want— and all those things, they're very amenable to those things. Uh, if we do a direct relationship.
Mm-hmm.
However— I will say I am concerned about their, um, scalability.
Oh.
So, you know, one of the things that they quoted was they can get, they're gonna be able to generate three gigawatts annually.
Right.
Right? They, they, they believe they can deliver that.
Mm-hmm.
And we're pretty skeptical [laughs]
[laughs]
... about that right now. Um, even if they can, right, you know, that would be divvied off between maybe many hyperscalers and co-locations and electrical integrators. So how much can they guarantee to us per year is concerning, right? Because again, right, if of that three gigawatts, you can only give me 100 megawatts—
Mm.
... that's maybe not worth it to me.
Mm.
Right? That's, uh, not a lot. That's, you know, maybe what? That's one third, one half of, well, half of one third, right?
Right. Right.
So I'm only getting 12% of what you're giving out to the entire market, which overall may be okay, again, if through a leasing agreement.
Mm-hmm. Mm-hmm.
Um, but, uh, I, I definitely would be, uh... That, that's what we're skeptical of right now. We're skeptical of if they can actually meet that scalability, and we want ways for them to prove it.
Mm.
Then, um, I remember that FTAI promised delivering like two to three gigawatt per year by year 2027. So—
Correct.
Would AWS—
Which would be like around the time they'll be fully qualified if we use them, right?
Oh.
So like if we, we start the process right now— by the time they're, they're like ready to go, it would be 2027.
Right. So, um, I guess AWS would like to see whether FTAI can, uh, successfully, like, have that two to three gigawatts of scale, and then, uh, like make the actual order?
Correct. Right. Because they're saying that their gas turbines can be 25 megawatts.
Right. Right.
Right?
Mm-hmm.
So 25 megawatts, that's if you do four of them, that's 100—
100.
... megawatts. So if I'm only getting 100 megawatts, that means I'm only getting four engines.
Mm-hmm.
That's not a lot for me.
Right. Right.
Right? I want more than just four. Uh, so yeah, so I think that for us, right, we want to see that, like, they can actually make that amount of value. Um, so that's gonna be work- that's gonna be one of the things that can have to work through that process.
Um, and then, like also for us, like the, what we wanna see, you know, but this is gonna be more so... This is not commercial terms, so this is more like a technical thing— that we, we'll have to discuss later. But, you know, they're, they're converting jet engines— right?
So, right, war jet engines, they're converting those to, uh, uh, engines for data center power generation. Um, we want to see if that is even feasible, right? So that they have to show that in the market, right? So that's one of the things for us is like, it's not like an official term, but, like, we want to see them do this on a smaller scale to other players as well, right?
Mm-hmm.
Which is kind of counterintuitive, right? Because you think, oh, maybe Amazon wants to use all of their, like, you know, get all of their two gigawatts.
Yeah.
But we don't, right? We want also to see if, you know, in, in a pra- from a practical standpoint— does this generate the amount of power that they think it can generate?
Right. Then, uh, based on the current discussion that you're having with FTAI, um—
Mm-hmm.
... like, is there a rough timeline that FTAI promised that they can show some, uh, like signals or, uh, hints of them, uh, them that they can, uh, like, show this—
Yeah.
... feasibility in terms of converting those engine, as well as like, uh, uh, showing that they have the scalability?
Yeah.
So, uh, they were thinking about, like, more so scalability right now. Like, I think that the, again, like... So I guess, like, the best way to think about it, first scalability. Can they scale? Um, can they show us, like, their production, all that?
Right.
That, that is Q4, early Q4 this year.
Ah.
Um, that they expect to be able to do that. Now, um, in terms of, like, actually showing it in the market, we expect to see that in Q1 2027-
Mm.
Where we actually see it in the market. And if we see that, then, like that, that was... Again, but that's more of a, on the technical end. That's when we start to get almost to the end of that qualification process.
Mm.
Um, so for, to answer your question of, like, Q4 for the, um-
Okay.
...actual scalability question, right?
Mm.
We wanna see the factories, we wanna see the production lines-
Right.
...stuff like that.
Ah.
Uh, and then Q1 for, like, actually being in the market, and they're actually, like, selling it to other players.
Mm. And when you are, like, um, asking some, like, process or some possibility of FTAI, like, uh, scaling their production facility as of now, like-
Mm-hmm.
...what, what kind of feedback it, uh, does FTAI, uh, give, give to you?
Yeah. So typically, uh, what they say is, like, "Hey, here, like, this was what used to be a aircraft engine production facility. It's now changing to a aeroderivative-
Right.
...production facility." And they show, like, their plans. They show, like, how many, uh, components that they're making for the different engines per, uh, per week. And they basically just give, they just give us, like, uh, rundowns of what their production capacity is like-
Mm-hmm.
...and things like that.
Mm. And I guess AWS is evaluating, like, other power engine solution vendors as well, including FTAI.
Oh, yeah. All the time.
Yes. And-
Of course, yeah.
Um, comparing, comparing, compared to, like, those other competitive-
Mm-hmm.
...uh, providers, how do you think of FTAI's, uh, like, um, arguments-
Their goals.
...or, like-
Yeah. So-
As of now.
...their, their pitch is different. Uh, you know, a lot of more legacy players, I guess, uh, so to speak, are... Right? They, they already can show, you know, that this is what they're doing, right? They already say, "Hey, I already make..." You know, like a, uh, a GE or a, uh, Mitsubishi, they already say, "Hey, I'm already making-
Right.
...these things," right? So when, usually when they qualify something-
Mm-hmm.
...they're just qualifying, like, a bigger, a, a bigger turbine.
Mm-hmm.
Or, you know, another good example would be, like, a, uh, there's a China, a Chinese-based company called, uh, uh, Weichai Power, right? So, you know, they're, they're gonna have, like, smaller engines. They're only five megawatts. But- They, they, you know, they already do it. They just have it in China.
So they have to prove that they can produce them in the US because we don't want to use China-based-
Mm-hmm.
...engines.
Mm-hmm.
But like they're, they're... Usually like the use cases of what they're presenting is a little bit different than what FTAI is doing, which is saying like I can even... Like, I can be a player at all. And so their, their approach is different, so I will say that. But I will also say that it's, it's not very data-driven like I would want it to be.
Oh.
Like I would want more data-driven, uh, information.
Mm-hmm.
When they talk about production capacity and scale, so I wish they could do that better. Um, and I w- and I want more maybe like tours of the factory and things like that-
Right.
...so I can see things. Uh, which that's something like the other, um, the other players do. FTAI, I feel [chuckles] that they're a little bit secretive right now.
[laughs] Oh.
And, and, uh, uh, and that's always a, for me, a little bit of a, uh, red flag.
Oh. 'Cause I was also curious whether FTAI actually showed their like the actual factory to the, to potential-
Just pictures.
...vendors. Ah, only pictures.
Yes.
Oh. But is it possible to like actually gauge their, um, like production increase, like-
Right.
...process o- through those pictures?
If they were more transparent it would be, right? You know, if they gave us more data, right? Or like even a baseline, right? 'Cause like when they give us data on their production, right, they do give us some data, right? But like I don't know what I'm comparing this to. They just say, "Hey, we increased, we're increasing our capacity by 25%. We're hiring these maybe more people. We're opening up these more lines." But like, okay, based on a baseline of what? Like, what is your comparison to? Like, are you comparing yourself to, you know, nothing? 'Cause you weren't doing any-
Oh.
...you know, data center things before.
Mm-hmm.
Are you basing your comparison based on when you were doing your airplane engines, which is not apples to apples comparison.
Oh.
Uh, or like, you know, they could at least compare itself maybe to other companies, right? They can say, "Hey, Mitsubishi's making these big 20, 10, 20 megawatt engines at a rate of X per year."
Mm-hmm.
"And we're gonna match that." Right? Like, so stuff like that is what I would want to see more of, and I don't think they're doing a, a good job of that. So like for me, like to be quite honest, right now it's very hard for me to gauge if they can meet our scale or like... Even for me at all, right? I don't-- I have no way of knowing of like, no good way of judging if they are, you know, are... If they're gonna actually meet those two gigawatt expectations.
Mm-hmm. Oh, interesting. Did, did, did FTAI actually like, um, suggest like visiting their factory so that they can-
No.
No? Oh. Did-
I want to. If they ever suggested it I would love to go.
Oh. Ha- has AWS actually like requested such a factory visit?
We, we haven't requested yet.
Uh-huh.
Uh, because, you know, we've... We started the process in February. It's been four months.
Oh.
And we're just really trying to narrow down the master service agreement. Like, like, do they agree to the terms that we want?
Mm-hmm.
So before I even board, like before I would ask to go to a facility, I want to hash that out. But like if they were to, on the other hand, if they were to ask us, we would definitely, um...
We would, we would, we would definitely do it. I would love to visit some of their facilities. Even if the facilities haven't been fully converted yet—
Right. Right.
...I would still love to see it.
Mm-hmm. Oh, and if you happen to visit their facility, can you also give a notice to the Tiro so that I can request a follow-up call to get a sense—
Yes.
...of like how they actually like advance their expansion capability?
Yes.
Oh, thank you. And also, are there any other like new entrants like FTAI that AWS is currently evaluating as a potential vendor, like besides FTAI?
Yes. So I mentioned one, Weichai Power—
Oh, the Chinese.
...which is a China company. But you know, their US arm is going to be called Power Solutions International, or PSI.
Oh.
But they're gonna make really smaller engines, right? Like three megawatt engines.
Oh.
So like they're gonna be mostly used for us for our backup power.
Oh.
Not for like primary power.
Oh.
Or even like, you know, for FTAI, we hope to use them as a bridge until we get connected to utility power. Where for PSI we're gonna be mostly just using them for like just backup power.
Mm-hmm.
But those are the... That's like a company that we're looking to qualify and use their natural gas turbines.
Of course, like I mentioned, the issue is that PSI is now owned by Weichai Power, which is a China-based company.
Mm-hmm.
And, uh, we always have some reservations about Chinese manufacturers, especially when it comes to natural gas engines.
Mm-hmm.
We typically... If we get from APAC, right, it's usually like Japan—
Mm-hmm.
...or Korea. And also, right, we are gonna be very cognizant of, like, tariffs, if we get it from—
Oh.
...uh, China. Even though—
Right, right.
...right now I'm not too concerned about that.
Mm.
But it could be a concern in the future.
Mm. Then, if AWS is currently considering FTAI as a bridge power solution—
Mm-hmm.
...I mean, does that mean that it is tailored for, like, near-term capacity shortages, like rather than a long-term strategy?
Well, no, I mean, right, so I guess, like, as a context.
Mm-hmm.
When I say bridge, this is gonna be a three to five-year bridge.
Oh.
Right? 'Cause that's how long it sometimes takes you to, for us to get connected to utility power.
Right. Right.
But that's always the long, the end goal, right?
Mm-hmm.
To get connected to utility power.
Mm-hmm.
But yeah, it would be a three to five year bridge.
Mm-hmm.
So I wouldn't call that short term.
Mm-hmm.
I would call it more midterm.
Mm-hmm. Mm-hmm. Mm. I see.
Then, if FTA successfully expands their capacity into two to three gigawatt per year by year 2027—
Mm-hmm.
...how much of a gigawatt power would AWS like willing to make an order? Like, if—
[laughs]
...if they successfully—
I, uh—
Like in a—
If I—
...in hundreds of—
Yeah.
...megawatt or, like, gigawatt—
If—
...size?
If, if I were... So of that three gigawatts, if I were to ask my vice president—
Mm-hmm.
...he would want me to get two gigawatts of it. [laughs] He would want us to get most, most of it.
Uh.
I just know that's not realistic, right?
Right.
'Cause we'd be competing with other players.
Uh.
But let's say, you know, this is hypothetical, right?
Right.
Let's say hypothetically speaking, they pass all of our inspection.
Uh-huh. Uh-huh.
We want to order from them directly.
Uh-huh.
And they're, like, perfect with everything in terms of engineering and scalability and scope. I would want to take two gigawatts.
Mm. And—
But I know that's, like, not realistic.
Uh.
But realistically speaking, I see us probably taking around 500 megawatts, which is probably like 25%.
Mm. Is it because, like, when you are talking with FTAI, it seems that FTAI is also in a conversation with other hyper—, other—
Oh, yeah.
...like, major hyperscalers as well—
Yes.
...and they are—
Which, which is a good—
Yeah.
...this is, it's a good and a bad thing, right?
Mm.
It's a good thing because, like, if they're talking to other hyperscalers and other co-locations, that means that will... That tells me that they are serious and also it shows that they would have to also prove it to these other players as well.
Mm.
So I mean, they're putting more and more into their R&Ds, right?
Mm.
So that's the good thing.
Mm.
And of course, but the bad thing is that now, of course, we'll be competing just like we compete with all the other players. That's why, again, that's why I mentioned, you know, one of the positives about FTAI is they're just saying that they're going to have really good lead times. But by the time 2027 rolls around, when we can actually get orders, it's probably gonna be long lead times [laughs] because all the other players are gonna put in orders as well.
Mm.
And you know what's gonna happen, as we all know, is that they're gonna say, "Hey, we can deliver three gigawatts."
Mm.
But they're gonna allow orders for, like, five gigawatts.
Mm.
And they're gonna put people in back order and things like that, right?
Mm.
So even if everything goes perfectly, I still don't foresee us getting, you know, everything that we want.
Mm. Then, do you know whether, like, a certain hyperscaler is more in, like, an, uh—
Yeah.
...in a more advanced—
Yeah, that I'm not sure of.
...process? Uh.
Right. I know we're probably pretty stringent, I will say, right? Like AWS with all, with our law team, our finance team, and the fact that, right, like for AWS, we're just not like a typical hyperscaler. We are more of a legacy hyperscaler.
Oh.
I'm sure in that, like we have not just AI ML, we have networking, we have storage racks, we have all different types of racks. While I think, like, for Meta and Google or Grok, you know, emerging, emerging hyperscaler—
Oh.
... they're probably more willing to put money into CapEx, maybe more willing to expedite these processes. However, I'm, I'm not sure.
Mm. Uh, does FTAI give some hints, uh-
No.
... regarding, like, how competitive, like, other hyperscalers want it or want make their-
N- not yet.
Not yet?
I think they'll, they'll probably start doing that more so as we move along-
Mm.
... in the qualification process. But as of right now, they have not.
Mm. And, um, regarding the pricing of their engines, uh, do you think that-
Mm-hmm.
... the price range that they are currently offering is in a, uh, like industry average level, or is more like, um, at a higher, at a higher level?
Well, it's still negotiable right now.
Uh-huh.
But, um, right, right, right now they're about $15 to $16 million per unit.
Mm-hmm.
Which is the one... Like, 'cause I, I'd like to think about that price per megawatt. So their price per megawatt is not, is, is, is right on par. Maybe a little bit higher right now.
Mm-hmm.
But, you know, like again, we haven't had-- we don't have set prices yet.
Ah.
So, uh, when we get the set prices, I, I would say, you know, I could give a better idea. But at least for what they... Like at a high level, what they showed us already, the ASP, uh, per megawatt is gonna be on par with like a GE or Mitsubishi if we buy direct.
Mm-hmm.
Now if we lease it, it's gonna be a little bit different, but you always, you know, we always pay more per megawatt if you lease.
Mm-hmm. I see. So it seems that as long as FTAI can successfully show their, uh, manufacturing, uh, ca- expansion capability, then-
Mm-hmm.
... hyperscalers would, uh, willingly like place an order to FTAI?
It, it, again, it also comes to the terms, right?
Ah, yeah, the terms. Mm-hmm.
I, I... The, the thing is gonna-- the big thing is gonna be like do we buy purchases, do we lease? And then if we wanna lease, then are they-- do they have, um... Are they willing to work with our electrical integrators, right? So that's gonna be the big thing. Um, but in terms of if everything else goes good, then yes. The only thing that would be the question is rent versus lease.
Mm-hmm.
Or excuse me, buy, buy versus lease.
Buy versus-
And if it's a lease-
Mm-hmm.
... then like they will have to take another step in working with our electrical integrators. Uh, not just for us, but like for other hyperscalers as well probably, you know, to work with their electrical integrators to make the purchase of the engine and then lease it to the hyperscaler-
Mm-hmm.
... or lease it to us.
Um, are you d- are you aware of like whether FTAI's current sales strategy is through only this direct purchase, uh, option? Or they're going, they are going for this, um, integrator, like through-- going through the electrical integrator, like sales strategy as well?
Yeah. I-- my, um... It, it, it really does depend on how, how they position their equipment. If they see, if they position these as backup generators, then these will be, uh, owned, right? 'Cause I'm gonna own a backup generator. But if they say, "Hey, these are for behind the meters. These are things that we want you to have long term"-
Mm-hmm.
... then that's when we say we would rather lease. Because right, like I can then, uh, get rid of that CapEx, right? I don't have any CapEx because I'm not buying it, right? So it, it, it really does depend on, um, how the structure would be in terms of the capacity structure, right?
Mm-hmm.
Is it gonna be something where we have energy as a service, which we will go through electrical integrator, or do we build our own microgrid, which I don't think we will. Uh, so like it depends on how we use the engine.
Mm-hmm.
If we use it as a backup, we will own it. If we use it as a bridge for our utility-
Mm-hmm.
... uh, we will, um, we will, uh, more than likely lease it.
Mm-hmm. I see. Then, uh, compared to those other OEM, uh, solutions and new entrants-
Mm-hmm.
... um, FTAI's, uh, advantage would be, uh, the speed, and also-
Yes.
... not coming from this Chinese, uh, like geo- geographical risks.
W- well, one other thing-
Mm-hmm.
... that is a potential, which we're still the early talks on, is that FTAI could potentially be their own electrical integrator. And, 'cause they own their own power-producing facilities, right?
Mm-hmm.
So what they can do is keep... You know, depending on where they are, right? They have a plant in Montreal, they have a plant in Miami. What they could potentially do is, um, like basically, uh, have us sign a PPA, which is a power purchase agreement.
Mm-hmm.
When they say, "Hey, this power is available to you-
Oh.
... and we can generate the power for you in our facility." And then they're kinda acting as an electrical integrator. So like we're leasing it from FTAI directly.
Mm-hmm.
So they will own the-- they will continue to own the engines themselves-
Mm-hmm.
... but then we're paying them to lease it from them.
Mm-hmm.
So we're leasing the engine directly from FTAI-
Mm-hmm.
... and they're helping us with the electrical integration. But that will take a little bit even longer-
Mm-hmm, mm-hmm.
... to get that to be pulled through. Um, so I, I, I, I would definitely say like direct ownership, um, is what we would probably have to do because it's faster.
Right.
Like ideally, I would want an indirect relationship-
Mm-hmm.
... but like more than likely, due to speed, I, I would probably have to buy the engine.
Mm-hmm.
And maybe one of the disadvantages that FTAI, uh, currently has would be some-- being somewhat secretive [chuckles] in terms of their, um, like scalability. Issue?
Yeah, at least for now.
Mm-hmm.
Right? And I, and I get it 'cause this is early, so right.
Right.
They, you know, they made a big claim, and they probably cannot back it up yet. So they probably said, "Hey, let's, let's wait until we can actually back it up before we start to put more information to the market."
Mm.
'Cause, you know, that would, you know, that can change their stock price. [chuckles]
Right. But, um, six months already passed since their announcement-
Yeah. [laughs]
-of that, uh. [laughs]
Yeah.
So I was curious whether from hyperscaler perspective they, uh, you are seeing some of those like, um, mo- movements and progress-
No.
... in terms of their-
Uh, I, I ha- I have not. Um-
Oh.
... even talking to some peers, they have not.
Oh.
So it maybe, um... You know, I don't have any, uh, I don't know anyone at Groq. I don't know anyone at, um, at Google.
Mm-hmm.
So maybe they have better pathways-
Mm-hmm.
... uh, to it. But I know at least, uh, from my other peers that, you know, hyperscalers like Azure or, uh, Meta- uh, yeah, it's, it's, it's pretty much the same where, uh, not... It's not necessarily being secretive versus just that, like, they haven't given us a lot of new information.
Information. Mm-hmm.
Then is it, uh, is it normal for like hyperscaler to enter this MSA, um, sta- stage, like process? Even though the vendor-
So this is like- this is where we would discover these things, right?
Oh.
When we go to... When we're entering these MSA discussions- it's when we, that's when we uncover a little bit more information.
Mm.
That's like, that's what this process, at least for, you know, here- that's what the process is for.
Oh.
Uh, to kind of get that, to get that discovery- uh, going. And, and again, like it's only been four months, but like we are working through the terms. And then like I mentioned, the second step we'll get even more information.
Mm-hmm.
Uh, 'cause we'll have to... We'll start to get them to do some type of sup- supply commit.
Oh. I see. Oh. Uh, thank you so much for your, um, information and insight. Um, it was very helpful. And if you happen to get any updates regarding, uh, FTA's like execution, um, I will highly appreciate if you give a notice to the plat- Tiga's platform so that we can arrange a follow-up call.
Yes. Yes. Uh, yeah, I might hear something new, uh, in Q3, so maybe in August, uh, I can, uh, let them know that I got some new information.
Okay, great. Oh, and for a last comment, are, are there any comments that you would like to add on that, uh, that I should be aware of regarding FTAI or- about this industry in general?
I, the- I think the only, I think the only thing would be, I would want anyone that's looking into this market to, to really, uh, dive deep into like how much the cost of natural gas will cost. Like, how much would be the cost of natural gas? Because that's like a big thing too, right? If, if the cost of natural gas increases, hyperscalers like us are gonna look for other solutions.
Mm.
Because the cost of the natural gas, it could be, will really drive up the cost. Which again, cost is not a big thing versus speed.
Mm-hmm.
But if there are any emerging things like fuel cells- and, um, batteries, uh, that means I, I would still, again, I would still use natural gas, but maybe I'm not using natural gas as much as a backup power anymore.
Mm-hmm.
And now I'm only using it to bridge the gap, which means that, hey, in six or seven years when I get on the utility- maybe I don't need natural gas turbines anymore.
Mm.
But of course, though, th- and that's really far out, but that would be my only other comment.
Mm. And, uh, speaking of the fuel cell solution, um- is, is Bloom Energy's solution one of the options that AWS is, um-
Yes, but only for backup power.
Oh, only for a backup power, but not yet a primary.
Yeah, like it's another bridge if I'm not on the utility. So really Bloom only works... We only, we only can use it really if we are already on a utility or we're already using natural gas turbine as our primary power.
Oh.
So fuel cells at least, at least right now- except in South Korea. In South Korea we do sometimes use fuel cells as primary power.
Right. Right.
At least in the US. In the US fuel cells are only backup power.
Mm-hmm. Mm-hmm. Mm. Is it because of the cost, cost, uh, component that is blocking-
No, the efficiency.
Efficiency.
Natural gas is just more efficient.
Uh-huh. Oh.
Um, and as we move to more and more powerful GPU, we want to be most efficient with our power as possible. And by efficiency I mean, uh, that like, uh, natural gas has a lower chance of going out, and natural gas, at least for right now, creates less thermal energy. So that means there's less thermal energy you have to dissipate from the, uh, server racks.
Mm.
So it's, it's just natural gas, uh, turbines are just more efficient and faster right now.
Mm-hmm. But then like other data centers like Oracle, they also announced that they would start utilizing this-
Yeah, fuel cell.
... fuel cell as- in a gigawatt like size.
Yes. 'Cause it is something that's emerging, right? I don't think that, um, I don't think that like it'll always be backup power.
Mm.
I think that like eventually we could potentially use it- as a primary power. It's just that like fuel cells are so nascent.
Oh.
Like right now I don't think anyone is running fuel cells as primary power except for markets that they have to.
Mm.
Like again, like in South Korea- or like congested markets like Tokyo.
Uh-huh.
You have to be fuel cell because there's not, there's not enough grid power to connect to.
Uh-huh. Uh-huh. I see. Ah. Well, thank you so much for your time today.
All right. Thank you.
Thank you.
Have a good rest of your day.
Have a good rest of your day too. Bye.
Bye. [phone dings]