MOJO

2K posts

MOJO

MOJO

@momo56808367

๐Ÿ‡ฐ๐Ÿ‡ท๐Ÿ‡จ๐Ÿ‡ฆ๐Ÿงก$BTC/$IREN/$IRE

Canada Katฤฑlฤฑm Nisan 2019
858 Takip Edilen266 Takipรงiler
์„ธ๋ฐ”๊ฒฝ โšก๏ธ
$IREN: Horizon 1 ๋‚ด๋ถ€ ๋‚ฉํ’ˆ @Deep_miners๊ฐ€ Childress ํ˜„์žฅ ์ง์›๋“ค์„ ์œ„ํ•œ ์ถ•ํ•˜ ๋งŒ์ฐฌ์˜ ์ธ์Šคํƒ€๊ทธ๋žจ ๊ฒŒ์‹œ๋ฌผ์„ ๋ฐœ๊ฒฌํ•œ ํ›Œ๋ฅญํ•œ ๋ฐœ๊ฒฌ์ž…๋‹ˆ๋‹ค. ์ง์›์ด ๊ตฌ์ฒด์ ์œผ๋กœ "Childress์—์„œ์˜ Horizon 1 ์™„๋ฃŒ"๋ฅผ ์œ„ํ•œ ๋งŒ์ฐฌ์ด๋ผ๊ณ  ๋ช…์‹œํ–ˆ์Šต๋‹ˆ๋‹ค. ์ด์ œ Microsoft์˜ ์ˆ˜๋ฝ ํ…Œ์ŠคํŠธ๋ฅผ ๊ธฐ๋‹ค๋ฆฌ๊ณ  ์žˆ์Œ ๋“œ๋””์–ด H1 ์ธ๋„๋˜๋Š”๊ตฐ์š” ๐Ÿ‘๐Ÿป๐Ÿ‘๐Ÿป H1 : ~7์›” ๋ง H2-4 : ์—ฐ๋ง ๊นŒ์ง€ Lets Go ๐Ÿ”ฅ
Jim Liu@jiahanjimliu

$IREN: Horizon 1 Internal Delivery Great find by @Deep_miners who found the Instagram post for the celebration dinner for site employees at Childress. Employee specifically stated dinner was for the โ€œCompletion of Horizon 1 in Childressโ€. Now awaiting Microsoft acceptance testing!

ํ•œ๊ตญ์–ด
4
0
36
3.5K
Frans Bakker
Frans Bakker@FransBakker9812ยท
I forgot to x3 the GPU count for 3 data halls per Horizon, oops. Ao that's 3x 6,336 GPUs, or 19,008 GB300s per Horizon ๐Ÿ˜ฌ๐Ÿ‘
English
1
3
59
4.7K
Frans Bakker
Frans Bakker@FransBakker9812ยท
$IREN $NBIS The biggest risk for investor appetite, for AI-related debt, are construction delays. Investors are doing their homework This is exactly what we do on this platform. Going the extra mile to see the construction progress, while flagging early signs of delays, if any.
English
11
13
267
39.8K
Frans Bakker
Frans Bakker@FransBakker9812ยท
Despite various setbacks, $NBIS has stated that they are still on schedule with their deployment for $MSFT. $IREN will deliver their first tranche of roughly 6,336 GB300s, to $MSFT in the coming days, and has repeatedly stated to be confident in delivering the remaining 3 tranches, by the end of this year. Within the scope of our research, we have discovered an innovation that will help @IREN_Ltd increase the pace of building out Horizon 3 and 4, towards the end of this year. If my estimates are correct, this technology could save more than 19,000 work hours, of highly scarce workforce, in each data hall. What this means building cadence, is yet to be determined, but I currently attribute a 90% probability to the completion of the entire contract in this calendar year. Asset-backed financing is now within reach. Once these AI cloud companies establish a track record of timely delivery to hyperscalers, investor appetite for investment-grade debt should remain strong.
English
17
13
201
11.7K
MOJO
MOJO@momo56808367ยท
@logeonli16 @joajyoa8503 ๋„ต ์•„๋‹ˆ์—์—ฌ ใ…Žใ…Ž
ํ•œ๊ตญ์–ด
0
0
0
24
ํˆฌ๊ณฐ
ํˆฌ๊ณฐ@joajyoa8503ยท
$IREN ์•ฝ๊ฐ„ ์ œ๊ฐ€ ์ค‘๊ฐ„ ์„ค๋ช…์„ ์ƒ๋žตํ•˜๋ฉด์„œ ์˜คํ•ด๊ฐ€ ์ƒ๊ธฐ๋Š” ๊ฒƒ ๊ฐ™์•„์„œ ๋ถ€๊ฐ€ ์„ค๋ช…์„ ์ข€ ํ•˜๋Š”๊ฒŒ ๋งž๋Š” ๊ฒƒ ๊ฐ™์Šต๋‹ˆ๋‹ค. ์ด๋ฒˆ ์‹ค์  ๋ฐœํ‘œ๋Š” 6์›”30์ผ ์ด์ „๊นŒ์ง€์˜ ๋ฐ์ดํ„ฐ์ธ๋ฐ, 7์›”19์ผ์— H1์ธ๋„๊ฐ€ ์™„๋ฃŒ๋˜๋Š” ์‹œ์ ์— H1์˜ ์˜จ์ „ํ•œ ๋งค์ถœ์— ๋Œ€ํ•œ ์ด์•ผ๊ธฐ๋Š” ์•„๋‹™๋‹ˆ๋‹ค. 2์›” ์‹ค์ ๋ฐœํ‘œ์—์„œ CCO Kent Drapper๊ฐ€ Q2์— MS๊ณ„์•ฝ ์ดˆ๊ธฐ ๋งค์ถœ commence๊ฐ€ ์žกํž๊ฑฐ๋ผ๋Š” ๋ฐœ์–ธ์„ ํ–ˆ๋Š”๋ฐ, ์—ฌ๊ธฐ์„œ Q2๋Š” ํšŒ๊ณ„์ƒ์˜ Q2๊ฐ€ ์•„๋‹Œ ์บ˜๋ฆฐ๋” ์ƒ์˜ Q2๋กœ ์ €๋Š” ๋ณด๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค. ๊ทธ๋Ÿฌ๋ฉด ์ด๋ฒˆ ์‹ค์ ๋ฐœํ‘œ์—์„œ MS ๊ณ„์•ฝ ์ดˆ๊ธฐ ๋งค์ถœ commence๊ฐ€ ์žกํž๊ฑฐ๋ผ๊ณ  ๋ณด๊ณ  ์žˆ๋Š” ๊ฒƒ์ธ๋ฐ, ์ €๋Š” H1์˜ ์ธ๋„๊ฐ€ 8์›” ํ˜น์€ ๋Šฆ์œผ๋ฉด 9์›”๋กœ ์ƒ๊ฐํ•˜๊ณ  ์žˆ์—ˆ๊ณ , ๊ทธ๋Ÿฌ๋ฉด ์ด๋ฒˆ ์‹ค์  ๋ฐœํ‘œ์—์„œ MS ๊ณ„์•ฝ ์ดˆ๊ธฐ ๋งค์ถœ commence๊ฐ€ ์œ ์˜๋ฏธํ•œ ์ˆซ์ž๋Š” ๋ณด์—ฌ์ฃผ์ง€ ๋ชปํ• ๊ฑฐ๋ผ๊ณ  ์ƒ๊ฐํ–ˆ์Šต๋‹ˆ๋‹ค. ๊ทผ๋ฐ 7์›”19์ผ์ด๋ฉด ์ œ ์˜ˆ์ƒ๋ณด๋‹ค H1์˜ ์ง„ํ–‰ ์ƒํ™ฉ์ด 1๋‹ฌ๊ฐ€๋Ÿ‰ ๋น ๋ฅธ ์†๋„๋ผ๋Š” ์ด์•ผ๊ธฐ์ธ๋ฐ, ๊ทธ ๋ง์€ MS ๊ณ„์•ฝ ์ดˆ๊ธฐ ๋งค์ถœ commence๊ฐ€ ์ œ๊ฐ€ ์ƒ๊ฐํ–ˆ๋˜ ๊ฒƒ๋ณด๋‹ค๋Š” ์œ ์˜๋ฏธํ•œ ์ˆซ์ž๋ฅผ ๋ณด์—ฌ์ค„ ์ˆ˜ ์žˆ์ง€ ์•Š์„๊นŒ? ํ•˜๋Š” ์˜๋ฏธ ์ž…๋‹ˆ๋‹ค. Q2์— MS ๊ณ„์•ฝ ์ดˆ๊ธฐ ๋งค์ถœ commence๊ฐ€ ์žกํž ๊ฒƒ ๊ฐ™๋‹ค๋Š” Drapper์˜ ์ด์•ผ๊ธฐ๋Š” ๊ฒฐ๊ตญ H1์ด Q2์— ์ž‘๋™์„ ์–ด๋А์ •๋„ ์‹œ์ž‘ํ•จ์œผ๋กœ์จ H1์œผ๋กœ๋ถ€ํ„ฐ ์ž‘๊ฒŒ๋‚˜๋งˆ ๋งค์ถœ์ด ์‹œ์ž‘๋  ๊ฒƒ์ด๋ผ๋Š” ์ด์•ผ๊ธฐ์ธ๋ฐ, H1์˜ ์ง„ํ–‰ ์ƒํ™ฉ์ด ๋น ๋ฅผ์ˆ˜๋ก ๊ทธ ๋งค์ถœ์˜ ํฌ๊ธฐ์— ๋ณ€ํ™”๊ฐ€ ์˜ฌ ๊ฒƒ์ด๋ผ๊ณ  ์ƒ๊ฐ๋˜๊ธฐ ๋•Œ๋ฌธ์ž…๋‹ˆ๋‹ค. H1์˜ ์™„์ „ํ•œ ๋งค์ถœ์— ๋Œ€ํ•ด์„œ๋Š” ์ด๋ฒˆ ์‹ค์  ๋ฐœํ‘œ๊ฐ€ ์•„๋‹Œ, 4๋ถ„๊ธฐ ์‹ค์ ๋ฐœํ‘œ์—์„œ ํ™•์ธ์ด ๋  ๊ฒƒ ๊ฐ™๊ณ , ๋งŒ์•ฝ H2-4๊นŒ์ง€์˜ ์ง„ํ–‰ ์ƒํ™ฉ์ด H1์ฒ˜๋Ÿผ ๋น ๋ฅด๋‹ค๋ฉด MS์—๊ฒŒ H2-4๊นŒ์ง€๋Š” ์—ฐ๋ง๊นŒ์ง€ ์ˆœ์ฐจ์  ์ธ๋„๋ผ๋Š” ์ ์„ ์ƒ๊ฐํ•ด๋ณด๋ฉด 4๋ถ„๊ธฐ์— ์ƒ๊ฐ๋ณด๋‹ค AI Cloud ๋งค์ถœ์— ํฐ ๋ณ€ํ™”๊ฐ€ ์ƒ๊ธธ๊ฑฐ๋ผ๊ณ  ๋ด…๋‹ˆ๋‹ค. 4๋ถ„๊ธฐ์— ํฐ ๋ณ€ํ™”๊ฐ€ ์—†๋”๋ผ๋„ ์ ์–ด๋„ ๋‚ด๋…„ ์ดˆ ์‹ค์  ๋ฐœํ‘œ์—์„œ๋Š” ๋ณด์ผ๊ฑฐ๋ผ๋Š”๊ฒŒ ์ œ ์ƒ๊ฐ์ด๊ตฌ์š”.
ํˆฌ๊ณฐ@joajyoa8503

$IREN ์ด ์†Œ์‹์„ ๋“ฃ๊ธฐ์ „๊นŒ์ง€๋Š” ๊ทธ๋ž˜๋„ 8์›” ๋Šฆ์œผ๋ฉด 9์›”์— H1์ธ๋„๋ฅผ ์ƒ๊ฐํ–ˆ์—ˆ๋Š”๋ฐ, ๋งŒ์•ฝ ์ •๋ง 7์›” 19์ผ์— H1์ด ์ธ๋„ ๋œ๋‹ค๋ฉด 8์›”์— ์žˆ์„ ์‹ค์  ๋ฐœํ‘œ์—์„œ AI Cloud ๋งค์ถœ์— ์œ ์˜๋ฏธํ•œ ๋ณ€ํ™”๋ฅผ ๋ณผ ์ˆ˜ ์žˆ์„ ๋“ฏ ํ•ฉ๋‹ˆ๋‹ค. ์ด ์†Œ์‹๋Œ€๋กœ๋งŒ ์ง„ํ–‰์ด ๋˜๋ฉด, ์•ฝ๊ฐ„์˜ ๊ธฐ๋Œ€์ด๊ธฐ๋„ ํ•˜๊ณ  ๋ฐ”๋žจ์ด๊ธฐ๋„ ํ•˜์ง€๋งŒ ์˜ˆ์ƒํ–ˆ๋˜ ๊ฒƒ๋ณด๋‹ค๋Š” IREN์˜ ์žฌํ‰๊ฐ€๊ฐ€ ๋น ๋ฅด๊ฒŒ ๋‹ค๊ฐ€์˜ค์ง€ ์•Š์„๊นŒ? ์กฐ์‹ฌ์Šค๋ ˆ ์ƒ๊ฐํ•˜๊ณ ๋Š” ์žˆ์Šต๋‹ˆ๋‹ค. ์ฃผ๊ฐ€์ ์œผ๋กœ ๋„ˆ๋ฌด ํฐ ๊ธฐ๋Œ€๋ฅผ ์•„์ง์€ ํ•˜๊ธฐ๊ฐ€ ์–ด๋ ค์šด ์ด์œ ๊ฐ€, ์ด๋ž€ ๋ฌธ์ œ๊ฐ€ ์ƒ๊ฐ๋ณด๋‹ค ์ž˜ ์•ˆํ’€๋ฆด๊ฑฐ ๊ฐ™์€ ๋ถ„์œ„๊ธฐ์ด๊ณ  ์•„์ง์€ ์‹œ์žฅ์ด AI์— ๋Œ€ํ•ด์„œ ์˜์‹ฌ์„ ํ’ˆ๊ณ  ์žˆ๋Š” ๋ถ„์œ„๊ธฐ๊ฐ€ ๋‚จ์•„์žˆ๊ธฐ๋•Œ๋ฌธ์ด๊ธฐ๋„ ํ•˜๊ณ , ์‹œ์žฅ์€ ์–ด๋–จ์ง€ ๋ชจ๋ฅด๊ฒ ์Šต๋‹ˆ๋‹ค๋งŒ, ํˆฌ์ž์ž๋“ค ์‚ฌ์ด์— ์ด ์†Œ์‹์ด ๋“ค๋ ค์™”๊ธฐ ๋•Œ๋ฌธ์— ์‚ฌ์‹ค "17์ผ์— ์ธ๋„๊ฐ€ ๋˜์ง€ ์•Š์•˜์„ ์‹œ" ๋ผ๋Š” ๋ฆฌ์Šคํฌ๊ฐ€ ์ƒ๊ฒจ๋ฒ„๋ฆฐ ์…ˆ์ด๊ธฐ๋„ ํ•ฉ๋‹ˆ๋‹ค. ๊ทธ๋ฆฌ๊ณ  H1์ด ์ธ๋„๊ฐ€ 17์ผ์— ๋˜์—ˆ๋‹ค ํ•œ๋“ค, ์‹œ์žฅ์€ ๋˜ ๋‹ค์‹œ H2-4๊นŒ์ง€๋ฅผ ๋ณผ ์ˆ˜๋„ ์žˆ๋‹ค๊ณ ๋„ ์ƒ๊ฐํ•˜๊ตฌ์š”. ๋ถ€๋”” ๋‚ด์ผ ์ข‹์€ ์†Œ์‹์ด ๋“ค๋ ค์˜ค๊ธธ ๋ฐ”๋ž„ ๋ฟ์ž…๋‹ˆ๋‹ค. $IREN bullish

ํ•œ๊ตญ์–ด
4
1
16
3.5K
MOJO
MOJO@momo56808367ยท
@logeonli16 @joajyoa8503 ๋„ต ๋ฐด์ฟ ๋ฒ„ ์‚ด์•„์š”
ํ•œ๊ตญ์–ด
2
0
0
60
๋กœ๊ฑด๋ฆฌ
๋กœ๊ฑด๋ฆฌ@logeonli16ยท
@momo56808367 @joajyoa8503 ์ €ํฌ๋ฐฉ์— ์บ๋‚˜๋‹ค์‚ฌ์‹œ๋Š”๋ถ„ ์žˆ๋Š”๋ฐ ์บ๋‚˜๋‹ค ์‚ฌ์…”์š”?
ํ•œ๊ตญ์–ด
1
0
0
55
๋กœ๊ฑด๋ฆฌ
๋กœ๊ฑด๋ฆฌ@logeonli16ยท
@momo56808367 @joajyoa8503 ์—ฌ๋Ÿฌ๋ถ„๋“ค ์ œ๊ฐ€ ์šด์˜ํ•˜๋Š” ์นดํ†ก ์˜คํ”ˆ์ฑ„ํŒ…๋ฐฉ์— ์ฐธ์—ฌํ•  ์ƒ๊ฐ ์žˆ์œผ์‹ ๋ถ„ ์žˆ์œผ์‹œ๋ฉด ๋งํ•ด์ฃผ์„ธ์š”.
ํ•œ๊ตญ์–ด
1
0
0
44
MOJO
MOJO@momo56808367ยท
@logeonli16 @joajyoa8503 ๊ณต์‹ ๊ฐ€์ด๋˜์Šค๊ฐ€ q4 ๊นŒ์ง€์ด๋‹ˆ 10~11 ์›”๊ฒฝ ๋”œ๋ฆฌ๋ฒ„๋ฆฌ ์˜ˆ์ƒํ•ฉ๋‹ˆ๋‹ค.
ํ•œ๊ตญ์–ด
1
0
1
54
๋กœ๊ฑด๋ฆฌ
๋กœ๊ฑด๋ฆฌ@logeonli16ยท
@joajyoa8503 h1์ด ์™„๋ฃŒ๋˜๋ฉด h2 h3 h4๋Š” ํ›จ์”ฉ๋” ๊ฐ€์†๋„๊ฐ€ ๋ถ™์„ ์ „๋ง ์ž…๋‹ˆ๋‹ค. ๋น ๋ฅด๋ฉด ์—ฐ๋ง๋‚ด์— h4๊ฐ€ ๋งˆ๋ฌด๋ฆฌ ๋ ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.
ํ•œ๊ตญ์–ด
2
0
3
283
MOJO
MOJO@momo56808367ยท
@zsdcxa13 ์ปดํ“จํŒ… ๋ถ€์กฑ์ด ํ•ต์‹ฌ.
ํ•œ๊ตญ์–ด
0
0
1
270
๊ฒŒ๋กœ๋ฐ
๊ฒŒ๋กœ๋ฐ@zsdcxa13ยท
๊ตฌ๊ธ€์€ ์›๋ž˜ ์ง„์งœ ๋ฏธ์นœ ์žฅ์ ์„ ๊ฐ€์ง€๊ณ  ์žˆ์Œ. TPU๋ผ๋Š” AI ์ „์šฉ ์นฉ์„ ์ง์ ‘ ๋งŒ๋“ค์—ˆ๊ณ , ๊ฒ€์ƒ‰ยท์œ ํŠœ๋ธŒยท์•ˆ๋“œ๋กœ์ด๋“œยท์ง€๋ฉ”์ผ๊นŒ์ง€ ์ „ ์„ธ๊ณ„ ๋ฐ์ดํ„ฐ๋ฅผ ๊ฑฐ์˜ ๋‹ค ์ฅ๊ณ  ์žˆ์Œ. ๊ทธ๋Ÿฐ๋ฐ ์ง€๊ธˆ ์ƒํ™ฉ์ด ์˜ ์ข‹์ง€ ์•Š๋‹ค๋Š” ๊ฑฐ์•ผ. ํŠธ์œ—์—์„œ ์ œ์ผ ์›ƒ๊ธด(์Šฌํ”ˆ) ๋ถ€๋ถ„์ด ์ด๊ฑฐ์•ผ. ๊ตฌ๊ธ€์ด ์ž๊ธฐ๋„ค๊ฐ€ ๋งŒ๋“  TPU๋ฅผ Anthropic(Claude ๋งŒ๋“  ํšŒ์‚ฌ)ํ•œํ…Œ ๋Œ€๋Ÿ‰์œผ๋กœ ํŒ”๊ณ  ์žˆ๋‹ค๋Š” ๊ฑฐ. ๊ทธ๋ž˜์„œ Claude๊ฐ€ ์ฝ”๋”ฉ ์ž˜ํ•˜๋Š” ๋ชจ๋ธ๋กœ ์˜ฌ๋ผ์„œ๊ณ , ๊ตฌ๊ธ€์€ ๊ทธ๊ฑธ ์ง€์ผœ๋งŒ ๋ณด๊ณ  ์žˆ์Œ. ๋‚ด๋ถ€๊ฐ€ ์ง„์งœ ๋ฌธ์ œ๋ผ๊ณ  ํ•จ. Gemini ํŒ€, Search ํŒ€, DeepMind ํŒ€ ๋“ฑ์ด ์ปดํ“จํŠธ ์ž์›์„ ๋†“๊ณ  ์„œ๋กœ ์‹ธ์šด๋‹ค๊ณ . ์ž๊ธฐ๋„ค๊ฐ€ ๋งŒ๋“  ์ข‹์€ ์นฉ์„ ๋‚ด๋ถ€์—์„œ ์ œ๋Œ€๋กœ ๋ชป ์“ฐ๊ณ  ์žˆ๋‹ค๋Š” ๋œป์ด์•ผ. ๊ฒŒ๋‹ค๊ฐ€ Gemini 3.5 Pro๋„ ๊ณ„์† ๋ฏธ๋ฃจ๊ณ  ์žˆ์Œ. ์˜ฌํ•ด Google I/O์—์„œ โ€œ๋‹ค์Œ ๋‹ฌ์— ๋‚˜์˜ฌ ๊ฑฐ๋‹คโ€๋ผ๊ณ  ํ–ˆ๋Š”๋ฐ, ์ง€๊ธˆ๋„ ์•ˆ ๋‚˜์™”์–ด. ์ฝ” ๋”ฉ ์„ฑ๋Šฅ์ด ๋ชฉํ‘œ์— ๋ชป ๋ฏธ์นœ๋‹ค๋Š” ์ด์œ ๋กœ ๋ฏธ๋ค„์ง€๊ณ  ์žˆ๋‹ค๊ณ  ํ•จ. ์—ฐ๊ตฌ์ž๋“ค๋„ ๊ณ„์† ๋น ์ ธ๋‚˜๊ฐ. ์šฐ์ˆ˜ ์ธ๋ ฅ๋“ค์ด Anthropic์ด๋‚˜ ๋‹ค๋ฅธ ๊ณณ์œผ๋กœ ์ด๋™ํ•˜๊ณ  ์žˆ๋‹ค๋Š” ์†Œ์‹์ด ๊ณ„์† ๋“ค๋ ค. ๊ทธ๋ฆฌ๊ณ  ์ค‘๊ตญ ์˜คํ”ˆ์†Œ์Šค ๋ชจ๋ธ๋“ค์€ ์ ์  ๋นจ๋ผ์ง€๊ณ  ์žˆ์–ด์„œ, ๋น„์‹ผ ๋ˆ ๋“ค์—ฌ ๋งŒ๋“  ๋ฏธ๊ตญ ๋ชจ๋ธ๋“ค์„ ๋”ฐ๋ผ์žก๋Š” ์ค‘์ด๋ผ๊ณ . ๊ฒฐ๋ก ์ ์œผ๋กœ ๊ตฌ๊ธ€์€ ๊ด‘์‚ฐ(๋ฐ์ดํ„ฐ) + ์ œ๋ จ์†Œ(TPU) + ์ฒ ๋„(์ธํ”„๋ผ) ๋ฅผ ๋‹ค ๊ฐ€์ง€๊ณ  ์žˆ๋Š”๋ฐ, ์ •์ž‘ ๋ฏธ๋ž˜๋ฅผ ๋งŒ๋“œ๋Š” ๊ฑด ๋‹ค๋ฅธ ํšŒ์‚ฌ๋“ค์ด๊ณ , ๊ตฌ๊ธ€์€ ๊ทธ๊ฑธ ๋นŒ๋ ค์ฃผ๋Š” ์ชฝ์œผ๋กœ ๊ฐ€๊ณ  ์žˆ๋‹ค๋Š” ๊ฑฐ์•ผ. โ€œGoogle is not running out of assets. It is running out of excuses.โ€ (์ž์‚ฐ์ด ๋ถ€์กฑํ•œ ๊ฒŒ ์•„๋‹ˆ๋ผ, ๋ณ€๋ช…๊ฑฐ๋ฆฌ๊ฐ€ ๋ถ€์กฑํ•ด์ง€๊ณ  ์žˆ๋‹ค) ๋งˆ์ง€๋ง‰์œผ๋กœ โ€œ๊ตฌ๊ธ€์ด ์ƒˆ๋กœ์šด IBM์ด ๋˜๊ณ  ์žˆ๋Š”๊ฐ€?โ€๋ผ๊ณ  ๋ฌป๋Š”๋ฐ, ํ•œ๋•Œ ์„ธ์ƒ์„ ์ง€๋ฐฐํ–ˆ์ง€๋งŒ ์ด์ œ๋Š” ์ ์  ๊ด€๋ฃŒ์ฃผ์˜์— ๋น ์ ธ์„œ ๊ธฐํšŒ๋ฅผ ๋†“์น˜๊ณ  ์žˆ๋‹ค๋Š” ๋‚ ์นด๋กœ์šด ์ง€์ ์ด์—ˆ์–ด. AI ํŒ์ด ์ ์  ์žฌ๋ฐŒ์–ด์ง€๋„ค. ๊ตฌ๊ธ€ vs Anthropic vs ์ค‘๊ตญ ๋ชจ๋ธ๋“ค ๊ตฌ๋„, ์–ด๋–ป๊ฒŒ ๋ ์ง€ ๊ณ„์† ์ง€์ผœ๋ณด๋Š” ์ค‘. #AI #๊ตฌ๊ธ€ #Gemini #Anthropic #TPU
Ben Pouladian@benitoz

Be me: Google. Invent the TPU. Own Search, YouTube, Android, Gmail, Maps, the cloud, the data and half the internetโ€™s distribution. Have every advantage God and the antitrust division accidentally allowed one company to accumulate. Somehow still create an internal Hunger Games where Gemini, Search, Gmail, Waymo, DeepMind, and Maps fight over compute. Sell up to one million TPUs to Anthropic so Claude can become the best coding model on the infrastructure you invented. Then promise them multiple additional gigawatts. Reportedly pay 2x market for SpaceX Blackwell capacity because apparently inventing the TPU did not solve the minor operational challenge of having enough compute. Bring in Blackstone with $5 billion to build a separate 500 MW TPU cloud so outsiders can rent even more of your chips. Watch top researchers leave. Delay Gemini 3.5 Pro for months because it still cannot hit internal coding targets. Meanwhile, open-weight Chinese models ship, improve and start embarrassing closed American models that cost billions more to train. Then host Google I/O and have your own AI Overview call it one of the most boring keynotes in company history. Your AI is not aligned. It is roasting management. Google owns the mine, the refinery, the railroad and the shovels. Anthropic gets the TPUs. SpaceX gets the premium. Blackstone gets the infrastructure business. Departing researchers get the upside. Open source gets the momentum. Google employees get: โ€œFY27 Gemini Compute Prioritization Alignment Sync โ€” Pre-Read Required.โ€ Google is not running out of assets. It is running out of excuses. Is Google becoming the new IBM? Not dead. Not irrelevant. Still enormous. Just slowly transforming from the company building the future into the company everyone else rents the future from.

ํ•œ๊ตญ์–ด
5
15
81
10.4K
๊ฐ€์˜ค๋ฆฌ์—ฐ
์–ด์ œ ์นœ๊ตฌ๋ž‘ ์–˜๊ธฐํ–ˆ๋Š”๋ฐ, ์ค‘์†Œ๊ธฐ์—…๋“ค ์ค‘์‹ฌ์œผ๋กœ ๋กœ์ปฌ AI ๊ฒฌ์ ์ด ๋Š˜๊ณ ์žˆ๋‹ค๊ณ . 2~3์ฒœ์ •๋„๋กœ ์“ธ ์ˆ˜ ์žˆ๋Š” ๋ชจ๋ธ ๊ตฌ์ถ•ํ•ด์„œ ๋„ฃ์–ด์ฃผ๋Š” ๋น„์ฆˆ๋‹ˆ์Šค๊ฐ€ ์น˜๊ณ ๋‚˜์˜ค๊ณ  ์žˆ๋‹ค๊ณ  ํ•˜๋”๋ผ. ๋‹ค๋“ค ๊ด€์‹ฌ์€ ์žˆ๋Š”๋ฐ ์•„์ง ์–ด๋–ป๊ฒŒ ํ•ด์•ผํ• ์ง€๋Š” ๋ชจ๋ฅด๋Š”. ๊ทธ๋ž˜๋„ ๋ญ๋ผ๋„ ํ•ด๋ณด๋ ค๋Š” ๋‹จ๊ณ„์ธ ๋“ฏ. ๊ธฐ์—…ํ–ฅ์‹œ์žฅ์œผ๋กœ ์†Œ๋งค๊ฐ€ ์—ด๋ฆฌ๋ฉด ๋ณผ๋งŒํ•˜๊ฒ ๋‹ค.
ํ•œ๊ตญ์–ด
6
29
116
10K
MOJO
MOJO@momo56808367ยท
@Sebagyeong ๋‚ด์ผ..!?
ํ•œ๊ตญ์–ด
2
0
2
115
MOJO
MOJO@momo56808367ยท
@streitbar1900 Itโ€™s a Gb300 not a gb3000,
English
1
0
0
28
MOJO
MOJO@momo56808367ยท
@heemiiiin000 0.5 ์‹œ๊ฐ„์€ 10 ์ง„๋ฒ•์„ ์ „์ œํ•˜์— ๋งํ•˜๋Š” ๊ฑฐ์ž„. 4.5 ์‹œ๊ฐ„์„ 4 ์‹œ๊ฐ„ 50 ๋ถ„์œผ๋กœ ๋งํ•˜๋Š” ๊ฑฐ๋Š” ๋ฉ์ฒญํ•œ ๊ฑฐ์ž„
ํ•œ๊ตญ์–ด
0
0
0
58
ํ˜œ๋ฏผ
ํ˜œ๋ฏผ@heemiiiin000ยท
4.5์‹œ๊ฐ„์€ 4์‹œ๊ฐ„ 30๋ถ„์ด๋‹ค, ์•„๋‹ˆ๋‹ค ๋…ผ์Ÿ ๊ทธ๋Ÿฌ๋ฉด 3.5๋…„์ด๋ฉด 3๋…„5๊ฐœ์›”์ด๊ณ  3.5๊ฐœ์›”์ด๋ฉด 3๊ฐœ์›” 5์ผ์ธ๊ฐ€?ใ…‹ใ…‹ ์—ฌ๋Ÿฌ๋ถ„์€ ์–ด๋–ป๊ฒŒ ์ƒ๊ฐํ•˜์‹œ๋‚˜์š”?
ํ˜œ๋ฏผ tweet media
ํ•œ๊ตญ์–ด
59
13
66
21.7K
MOJO
MOJO@momo56808367ยท
@TBJorry ์กด๋ง›์ž„.. ๋‹ค์‚ฌ๋ž‘ ์น˜ํ‚จ ๊ทธ๋ฆฝ๋„ค
ํ•œ๊ตญ์–ด
1
0
1
33
์ชผ๋ฆฌ(JJory)
์ชผ๋ฆฌ(JJory)@TBJorryยท
1์ธ8์น˜ํ‚จ ๋“ค์–ด๋ณด์…จ๋‚˜์š”.....? ์ต์‚ฐ์˜ ๋‹ค์‚ฌ๋ž‘ ์น˜ํ‚จ์ธ๋ฐ, ์„œ์šธ ๊ณต๋•์ ์ด๋ผ๊ณ  ํ•จ. ์ต์‚ฐ์ด ์ง„์งœ ์ฐ์ด๋ผ๋Š”๋ฐ, ์ง„์งœ ๊ถ๊ธˆํ•˜๋„ค์š” ใ…‹ใ…‹ใ…‹ใ…‹
์ชผ๋ฆฌ(JJory) tweet media
ํ•œ๊ตญ์–ด
20
6
38
5.1K
๐€๐ ๐ซ๐ข๐ฉ๐ฉ๐š ๐ˆ๐ง๐ฏ๐ž๐ฌ๐ญ๐ฆ๐ž๐ง๐ญ๐ฌ
Reflecting on $IREN Over the last couple of days I spoke with multiple people in close contact with $IREN's management team, including investors who attended the RAISE Summit this week. Given the insights I've gathered, I think it's an appropriate time to reflect on $IREN and share my latest thoughts. It's no secret that $IREN has been somewhat slow on the commercial side, at least relative to the likes of $NBIS and $CRWV. I for one thought we'd have seen a Sweetwater deal by now, let alone substantial parts of the remaining Childress capacity pre-contracted. So what's stopping $IREN from signing these multi-hundred MW deals? In short, nothing is really "stopping" them. It comes down more to a few factors shaping their decision to hold off where other cloud providers perhaps wouldn't. Based on management's comments both on and off camera, I can confidently say demand truly isn't the issue. Cloud capacity in this market is sparse and supply can't keep up. In fact, I've heard $IREN could easily sell out 100% of its 2027 capacity today if it wanted to. The catch is that selling capacity which won't come online for another 6, 9, or 12 months yields significantly less than capacity arriving sooner. Customers want capacity today, and they're willing to pay a substantial premium for it. So while selling far into the future might prop up the stock, commercially it may not be the most prudent strategy in this environment. That dynamic can obviously shift over time, but given how far supply sits behind demand, it won't change overnight, and as it stands, holding off as long as possible yields better long-term returns. Not only do returns shrink the further out you pre-contract, but the available buyer pool shrinks with it. Selling capacity well into the future means gatekeeping much of the smaller, higher-margin clientele while mostly attracting the lower-paying hyperscalers. As we know, $IREN is increasingly moving up the stack, effectively cutting out the middle-man that hyperscalers represent, as evident in their recent Mirantis acquisition. On that note, $IREN apparently has multiple LOIs and customer commitments for high-margin managed cloud services set to take effect once the Mirantis deal closes over the coming weeks. I've now also heard several times that $IREN takes customer selection and contract structure extremely seriously. Creditworthiness matters, but management also wants clients that can scale their compute demand substantially as $IREN ramps capacity. The only near-term downside is that this due diligence takes time, yet the longer-term advantages of the approach are obvious. Beyond contract timing and customer selection, I believe some of it also comes down to operational reasons. We know the 1.4 GW Sweetwater campus is earmarked for the upcoming VR200 (Rubin) capacity, whose supply won't ramp until late this year into early next. That partly explains why the site isn't up and running already, since all they could lease out right now would be current Blackwell generation. The flip side is that $IREN could simply build "Horizon-style" capacity at Sweetwater, the same style they're currently developing at Childress, since those facilities are fully capable of housing next-gen Rubins, and have them ready by early next year, right as NVIDIA fully ramps Rubin production. And while $IREN is already doing foundation work at Sweetwater, it could still easily take another 3-4 quarters before we see operational capacity there. So what's the holdup? I believe a major reason for the slow ramp at Sweetwater is that they want to implement lessons learned from their Horizon build-outs at Childress, making the Sweetwater process more efficient, less costly, and thus more economical. Here I want to give a big shoutout to my friend @FransBakker9812, who found that $IREN has recently developed proprietary methods to make elements of the construction process significantly more streamlined, saving time and cost across all future liquid-cooled builds. He shared more specifics on that with his "Researchโ€ and โ€œFoundingโ€ subscription tiers, which I recommend checking out. I firmly believe what some might see as a relatively slow ramp, given $IREN's starting position, is management's way of doing things right. Start with the first liquid-cooled buildouts in Horizons 1-4, implement lessons from one Horizon batch to the next, then apply the full set of process and workflow improvements at Sweetwater. This closely mirrors what $IREN has always done since its mining era, when it started small and progressively scaled its construction operations in both size and speed. A true construction flywheel. Interestingly, I've just heard that $IREN plans to develop Sweetwater 1, Sweetwater 2, and the 1.6 GW Oklahoma site in parallel over the coming years. That shows just how exponential their construction ramp really is. In short, I believe holding out on the next wave of contracts comes down to a few factors: 1) Signing well ahead of commissioning means giving up pricing upside and attracting only a small subset of clients. 2) Customer selection and contract structure are a big part of $IREN's long-term strategy. It takes more time than simply selling to the highest bidder, but should build stronger customer relationships over the long run. 3) Scaling construction in a controlled manner, carrying critical lessons from current builds into the next. Slow start, exponential growth curve. None of this means we won't see any deals this year, but it does add color on why commercial progress on closing deals has been slower than many of us expected. As for deal activity and my current expectations there, it helps to step back and consider how $IREN's near-term capacity is structured. We should expect the 50k B300 units $IREN procured back in March to be fully contracted and installed by year-end, roughly 33k at Mackenzie and another ~17k at Childress. Apparently first deliveries for Mackenzie have already arrived and are being installed. Given this progress, I'd expect $IREN to announce having contracted substantial parts of these air-cooled Blackwells by August earnings at the latest. This is the low-hanging fruit. And worth noting, since $IREN first gave ARR guidance for that capacity, GPU rates across the board have moved up substantially. If they sign anything close to what they landed with the 60 MW NVIDIA deal, their year-end guidance of $3.7B should climb to at least $3.9-$4.1B. Beyond this, there's plenty of 2027 capacity that could get contracted later this year, including 190 MW of air-cooled capacity at Childress, 30 MW at Canal Flats, 150 MW of liquid-cooled Horizon 5-6, and 300 MW of liquid-cooled capacity at Sweetwater 1. We don't have guidance on when this capacity comes online next year or what the ramp schedule looks like, but since liquid-cooled greenfield development takes longer than retrofitting existing air-cooled buildings (currently mining BTC), I'd expect the remaining 220 MW of air-cooled capacity to come online within the first couple of quarters of 2027. For that reason, I think the odds those few hundred MW get pre-contracted later this year are relatively high. The trickier part is the 150 MW of Horizons 5-6 and the 300 MW of liquid-cooled Sweetwater capacity. I think there's a decent shot at least one of the two gets pre-contracted in 2026, especially if it's for a hyperscaler or a frontier lab, which are far more inclined to sign a few quarters ahead. Either way, it's just a matter of time until contracts start flowing. It's clear to me that $IREN is playing the long game and isn't compromising long-term upside for short-term euphoria in the share price. As a long-term investor, I fully support that. I do wish, however, that $IREN were a bit more open about strategy and roadmap. It's obvious they're holding their cards close to the chest, but I find management has been overly vague on strategy. It takes investors like me piecing the puzzle together to make sense of how $IREN plans to scale into the next hyperscaler. Ironically, management does share a fair bit of interesting and useful information if you get the chance to meet them in person, yet on earnings calls they come across as overly reserved. That said, the future looks bright, and I have no reason to get overly concerned about disappointing price action. With a bit of luck we're in for a string of positive catalysts, starting with the Horizon 1 handoff in a couple of weeks. I also want to take a moment to thank @OMCapitalGroup, who did an excellent job gathering information and insights while attending RAISE this week. If it weren't for his work, I wouldn't be nearly as informed, so big props to him for taking the time to travel all the way to Paris for $IREN due diligence and then going out of his way to keep me updated with everything he picked up, even putting some of my own questions to management directly. He's relatively new to X, but he told me he's going to start posting shortly and jump into Frans' spaces more often. Do me a favor and give this fella a follow. Have a good one, cheers! โœŒ๏ธ Thumbnail Credit (enhanced version): @AndyDTrades
๐€๐ ๐ซ๐ข๐ฉ๐ฉ๐š ๐ˆ๐ง๐ฏ๐ž๐ฌ๐ญ๐ฆ๐ž๐ง๐ญ๐ฌ tweet media
English
155
135
1K
174.5K
The Fundamentalist
The Fundamentalist@Funmentalistยท
$NBIS $29B ARR in July 2027? No way!!! Wait a minute, I might have calculated it wrong the first time... so Marc says: "Next 13 months, we are looking at tens of billions." So the next 13 months mean until July 2027 (if we count July 2026 as well) Letโ€™s say they will do $3B in the rest of the year, which would already be above the 2026 guidance as we have July already (Q2 ends in June, and they already made $400M in Q1 with the upper bound of guidance at $3.4B) We take the lowest "tens of billions" possible, which is $20B. They would have to generate $17B in the first 7 months of 2027 There is just no way that would put us at $2.42B/month on average $2.42B x 12 = $29.04B That would mean $NBIS would achieve $29.04B in ARR by July 2027 Obviously, as they ramp up their capacity, if they did $17B during those 7 months, the July revenue would be much higher than $2.42B, implying higher ARR than $30B I got around $25B ARR by EOY 2027 in my model Marc must have been a little euphoric and calculated it wrong; there is simply no way, and I think it would need a huge increase in prices with their capacity being available in these 13 months. And even then, it makes no sense. This number wonโ€™t let me sleep, absurd...
Molly Oโ€™Shea@MollySOShea

Marc Boroditsky (@marcboroditsky), Chief Revenue Officer of @nebiusai ( $NBIS ), says the next phase of AI is shifting from.. 'What are AI natives building?' to 'When are enterprises adopting?' "Lots of discussion about the CapEx, power & stock prices, which I think, at the end of the day, is a bit of a distraction because this is a marathon that we're running." "Actually being able to drive that enterprise adoption, that's the critical next stage for the entire industry." On the data center build-out: "The biggest challenge that many of our competitors are facing is being single-threaded, which is strange when you're in a multi-threaded parallel processing industry." Looking ahead.. "The next 12 months in AI is an eternity."

English
7
9
110
37.2K
MOJO
MOJO@momo56808367ยท
@Sebagyeong @two_god_ H1 ์„ธ๊ฐœ ๋นŒ๋”ฉ ์ค‘ 1 ๊ฐœ ์ •๋„๋Š” 6์›” ๋งค์ถœ์— ์ฐํž ๊ฐ€๋Šฅ์„ฑ๋„ ์žˆ๋‹ค๋ผ๋Š” ์นด๋”๋ผ๊ฐ€ ์žˆ๋”๊ตฐ์š” ใ…Žใ…Ž
ํ•œ๊ตญ์–ด
1
0
1
38
์„ธ๋ฐ”๊ฒฝ โšก๏ธ
@two_god_ h1์ด ์•ˆ์ •์ ์œผ๋กœ 7/์ค‘์— ๋‚˜์˜จ๋‹ค๋Š”๊ฒŒ ๋„ˆ๋ฌด ์ข‹๋„ค์š” ใ…‹ใ…‹ใ…‹ใ…‹ ์ด๋ ‡๊ฒŒ๋˜๋ฉด 8์›” ์‹ค๋ฐœ์—๋„ ๋ฐ˜์˜๋ ์ˆ˜์žˆ์„์ง€
ํ•œ๊ตญ์–ด
1
0
2
153
์„ธ๋ฐ”๊ฒฝ โšก๏ธ
$IREN ์ด๋ฒˆ RAISE Day์—์„œ ๋‚˜์˜จ ๋‚ด์šฉ๋“ค์ด ๋„ˆ๋ฌด ํก์กฑ์Šค๋Ÿฌ์›Œ์„œ ์ƒ์„ธํ•˜๊ฒŒ ์ •๋ฆฌํ•ด๋ณด๊ฒ ์Šต๋‹ˆ๋‹ค. 1/ ๊ฐ€์žฅ ์ค‘์š”ํ•œ ๊ฒƒ์€ $MSFT ๊ฐ€ ํ˜„์žฌ $IREN ๊ณผ์˜ ๊ณ„์•ฝ์— ๋งŒ์กฑํ•œ๋‹ค๋Š”์ . AI ์ธํ”„๋ผ ์‚ฌ์—…์€ ํ•œ ๋ฒˆ ๊ณ ๊ฐ์„ ํ™•๋ณดํ•˜๋Š” ๊ฒƒ๋ณด๋‹ค ์ฆ์„ค์ด ํ›จ์”ฌ ์ค‘์š”ํ•œ๋ฐ, MS๊ฐ€ ๋งŒ์กฑํ•œ๋‹ค๋Š” ๊ฒƒ์€ ์ถ”๊ฐ€ ๊ณ„์•ฝ ๊ฐ€๋Šฅ์„ฑ์„ ๋†’์—ฌ์ฃผ๋Š” ๊ฐ€์žฅ ๊ฐ•๋ ฅํ•œ ๋ ˆํผ๋Ÿฐ์Šค๊ฐ€ ๋ฉ๋‹ˆ๋‹ค. ๋‚จ์€ H5-8๋„ ๋งˆ์†Œ๊ฐ€ ๊ฐ€์ ธ๊ฐ€๋Š”๊ฑธ๊นŒ์š”? 2/ $NVIDIA ์™€ ๊ฑฐ์˜ ๋งค์ผ ํ˜‘์—…ํ•˜๊ณ  ์žˆ๋‹ค๋Š” ์ . ๋ชจ๋‘ ์•„์‹œ๋Š”๊ฒƒ์ฒ˜๋Ÿผ ์š”์ฆ˜ AI๋ฐฉ์„ ๋ฆฌ๋”ฉํ•˜๋Š”๊ณณ์€ ์—”๋น„๋””์•„์ž…๋‹ˆ๋‹ค. ๋ฐ์„ผ์—์„œ๋„ NVIDIA์™€ ์–ผ๋งˆ๋‚˜ ์ž˜ ํ˜‘์—…ํ•˜๋Š”์ง€๊ฐ€ ํ•˜๋‚˜์˜ ๊ฒฝ์Ÿ๋ ฅ์ด์ง€์š”. GB300 Exemplar Cloud ์ธ์ฆ, Vera Rubin ํ…Œ์ŠคํŠธ ์ฐธ์—ฌ ๋“ฑ์€ ๋‹จ์ˆœ ๊ณ ๊ฐ์‚ฌ๊ฐ€ ์•„๋‹ˆ๋ผ NVIDIA์˜ ์ฐจ์„ธ๋Œ€ ํ”Œ๋žซํผ์— ๊ฐ€์žฅ ๋จผ์ € ํฌํ•จ๋˜์—ˆ๋‹ค๋Š” ์˜๋ฏธ๋กœ ํ•ด์„ํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค. 3/ ๊ณ„์•ฝ ์กฐ๊ฑด. $IREN์€ ์ด๋ฒˆ ๊ณ„์•ฝ์ด 2๊ฐœ์›” ์ „๋ณด๋‹ค ํ›จ์”ฌ ์ข‹์•„์กŒ๋‹ค๊ณ  ์–ธ๊ธ‰ํ–ˆ๊ณ , ์„ ์ง€๊ธ‰ ๋น„์œจ๋„ 25~40%, ์ผ๋ถ€๋Š” 100%๊นŒ์ง€ ์ œ์•ˆ๋ฐ›๊ณ  ์žˆ๋‹ค๊ณ  ์–ธ๊ธ‰ํ•˜์˜€์ง€์š”. ๋งˆ์†Œ๋ฅผ ํ†ตํ•ด์„œ 20% ๋ฐ›์€๊ฑฐ ์น˜๊ณ ๋Š” ์„ ์ง€๊ธ‰๊ธˆ์ด ์˜ฌ๋ผ๊ฐ€๊ณ  ๊ทธ๋ฅผ ํ†ตํ•ด์„œ ์ž๊ธˆํ™•๋ณด๊ฐ€ ํ›จ์”ฌ ์šฉ์ดํ•ด ์ง„๋‹ค๋Š” ๋œป์ž…๋‹ˆ๋‹ค. ์ด๋Š” ์—ฌ์ „ํžˆ AI GPU ์ˆ˜์š”๊ฐ€ ๊ณต๊ธ‰๋ณด๋‹ค ๊ฐ•ํ•˜๋‹ค๋Š” ๋œป์ด๋ฉฐ, $IREN ์˜ ํ˜‘์ƒ๋ ฅ์ด ๊ณ„์† ์˜ฌ๋ผ๊ฐ€๊ณ  ์žˆ๋‹ค๊ณ  ์ฝํžˆ๋„ค์š”. 5/ Horizon 1, H2 ์‹œ์šด์ „, Sweetwater ํ™•์žฅ. ๊ฒฐ๊ตญ ๋ฐ์ดํ„ฐ์„ผํ„ฐ ๊ธฐ์—…์€ ์ „๋ ฅ์ด ์‹ค์ œ๋กœ ๋“ค์–ด์˜ค๊ณ  ๊ณ ๊ฐ์„ ์œ ์น˜ํ•˜์—ฌ ์‚ฌ์šฉํ•˜๋Š”๊ฒƒ์ด ์ค‘์š”ํ•ฉ๋‹ˆ๋‹ค. ์ด๋ฒˆ ๋ฐœํ‘œ๋Š” Horizon ์ง„ํ–‰์ƒํ™ฉ์ด ์ˆœ์กฐ๋กญ๋‹ค๋Š” ๋œป์œผ๋กœ ํ•ด์„ํ•˜์˜€์Šต๋‹ˆ๋‹ค.
Lando@LandoInvests

RAISE Day 1 Summary $IREN: โ€ฃ Horizon 1 handover confirmed 19th July โ€ฃ Commisioning H2 โ€ฃ MSFT very happy with IREN so far, very hands off. Great partners. โ€ฃ IREN on call with NVDA basically everyday โ€ฃ Deals signed today are even better than 2 months ago. โ€ฃ Pre payments sweetspot now between 25-40 roughly. Some even offering 100% โ€ฃ IREN GB300 NVIDIA EXEMPLAR Cloud certified. Likely announced next week. Only THREE have this right now. โ€ฃ IREN selected with small number for VERA rubin for testing VERY SOON. Unofficially late EOY mass production. Likely early 2027 โ€ฃ H5-H6, Sweetwater 1 Vera Rubin โ€ฃ Mirantis deal closing this month is the aim โ€ฃ Semi conductor shortage micron $MU seeing into 2031 h/t: @FinGigawatt x.com/LandoInvests/sโ€ฆ

ํ•œ๊ตญ์–ด
4
4
40
4.8K
๊ฒŒ๋กœ๋ฐ
๊ฒŒ๋กœ๋ฐ@zsdcxa13ยท
์š”์ƒˆ ๋ฉ”ํƒ€(Meta) ์†Œ์‹ ๋•Œ๋ฌธ์— ์ฆ์‹œ ์ง„์งœ ํ•ซํ–ˆ์ž–์•„. ๋‹ค๋“ค AI ๋ฒ„๋ธ”์ด๋„ค, ์ฒœ๋ฌธํ•™์ ์ธ ๋ˆ ์“ฐ๊ณ  ํšŒ์ˆ˜ ๋ชป ํ•˜๋„ค ๋ง ๋งŽ์•˜๋Š”๋ฐ, ์ด๋ฒˆ ์ €์ปค๋ฒ„๊ทธ ์ธํ„ฐ๋ทฐ ๋ณด๋‹ˆ๊นŒ ํŒ๋„๊ฐ€ ์™„์ „ํžˆ ๋ฐ”๋€” ๊ฒƒ ๊ฐ™์Œ. ๋”ฑ ๋ณด๊ณ  ํ™•์‹ ํ–ˆ์–ด. AI ๋ฒ„๋ธ” ์ „ํ˜€ ์•„๋‹ˆ๊ณ  ์ด์ œ ์ง„์งœ '์ธํ”„๋ผ ์ˆ˜์ตํ™”'์˜ ์„œ๋ง‰์ด ์˜ฌ๋ž๋‹ค๋Š” ๊ฑธ. M7์˜ CapEx(์ž๋ณธ ์ง€์ถœ) ํˆฌ์ž๋Š” ์•ž์œผ๋กœ ๊บพ์ด๊ธฐ๋Š”์ปค๋…• ํ•œ์ฐธ ๋” ์งˆ์ฃผํ•  ์ˆ˜๋ฐ–์— ์—†๋‹ค. ํ•ต์‹ฌ์€ ๋ฉ”ํƒ€๊ฐ€ ๊ทธ๋™์•ˆ ์ˆ˜๋ฐฑ์–ต ๋‹ฌ๋Ÿฌ ๋“ค์—ฌ ๊ตฌ์ถ•ํ•œ AI ์ธํ”„๋ผ(GPU ๋“ฑ)์˜ ์ผ๋ถ€๋ฅผ ์™ธ๋ถ€์— ์ž„๋Œ€ํ•˜๋Š” 'ํด๋ผ์šฐ๋“œ ๋น„์ฆˆ๋‹ˆ์Šค'๋ฅผ ๊ฒ€ํ†  ์ค‘์ด๋ผ๋Š” ๊ฑฐ์•ผ. ์™ธ๋ถ€์—์„œ "๋ˆ ์—„์ฒญ ํฌ๊ฒŒ ์ค„ ํ…Œ๋‹ˆ ์ปดํ“จํŒ… ์ž์› ์ข€ ๋นŒ๋ ค๋‹ฌ๋ผ"๊ณ  ์ œ์•ˆํ•˜๋Š” ๊ธˆ์•ก์ด ์›Œ๋‚™ ์„ธ์„œ, ๋ฉ”ํƒ€ ๋‚ด๋ถ€์—์„œ๋งŒ ์“ฐ๋Š” ๊ฒƒ๋ณด๋‹ค ์ผ๋ถ€ ๋นŒ๋ ค์ฃผ๋Š” ๊ฒŒ ์žฌ๋ฌด์ ์œผ๋กœ ํ›จ์”ฌ ํ•ฉ๋ฆฌ์ ์ด๋ผ๋Š” ๊ณ„์‚ฐ์ด ์„  ๊ฑฐ์ง€. ์ตœ๊ทผ ์ตœ์‹  AI ๋ชจ๋ธ์ธ '๋ฎค์ฆˆ ์ŠคํŒŒํฌ 1.1' API ์œ ๋ฃŒํ™” ์‹œ์ž‘ํ•œ ๊ฒƒ๋„ ๊ทธ ์ฒซ ๋‹จ์ถ”๊ณ . ์ด๊ฒŒ ์™œ ์†Œ๋ฆ„ ๋‹๋Š” ํฌ์ธํŠธ๋ƒ๋ฉด, ๊ทธ๋™์•ˆ ์‹œ์žฅ(์›”๊ฐ€)์ด ๋น…ํ…Œํฌ๋“ค ๋ณด๊ณ  "์•ผ, ๋ˆ๋งŒ ์—„์ฒญ ์“ฐ๊ณ  ์–ธ์ œ ๋ˆ ๋ฒŒ๋ž˜?"๋ผ๋ฉฐ ๋ฐœ๋ชฉ ์žก์•˜์ž–์•„? ๊ทผ๋ฐ ์ด์ œ ๊ทธ AI ์ธํ”„๋ผ๊ฐ€ ๋‹จ์ˆœํ•œ '๋น„์šฉ ๋ฉ์–ด๋ฆฌ(Cost Center)'๊ฐ€ ์•„๋‹ˆ๋ผ, ๊ณง๋ฐ”๋กœ ํ˜„๊ธˆ์ด ๊ฝ‚ํžˆ๋Š” ํ™•์‹คํ•œ ์บ์‹œ์นด์šฐ(Revenue Center)' ๋กœ ์ฒด์งˆ์ด ๋ฐ”๋€Œ๋Š” ๊ฑฐ์•ผ. ํˆฌ์žํ•œ ์ธํ”„๋ผ๋กœ ๋‹น์žฅ ๋ˆ์„ ๋ฒŒ๊ธฐ ์‹œ์ž‘ํ•˜๋‹ˆ๊นŒ ์žฌ๋ฌด์  ๋ถ€๋‹ด์ด ํ™• ์ค„์–ด๋“ค ์ˆ˜๋ฐ–์— ์—†์Œ. ์‹ค์ œ๋กœ ์ผ๋ก  ๋จธ์Šคํฌ์˜ xAI ๋ฐ์ดํ„ฐ์„ผํ„ฐ๊ฐ€ ๊ฒฝ์Ÿ์‚ฌ์ธ ์•คํŠธ๋กœํ”ฝ์ด๋‚˜ ๊ตฌ๊ธ€ํ•œํ…Œ ์ธํ”„๋ผ ํ†ต์งธ๋กœ ๋นŒ๋ ค์ฃผ๋ฉด์„œ ์—„์ฒญ๋‚œ ์›ƒ๋ˆ(ํ”„๋ฆฌ๋ฏธ์—„) ์ฑ™๊ธด ๋ชจ๋ธ์„ ์ €์ปค๋ฒ„๊ทธ๊ฐ€ ์ง์ ‘ ์–ธ๊ธ‰ํ•˜๋”๋ผ๊ณ . ์ž๊ธฐ๋“ค๋„ ๊ทธ๋Ÿฐ ์ œ์•ˆ ๋งŽ์ด ๋ฐ›๊ณ  ์žˆ์–ด์„œ ๋”ฐ์ ธ๋ณด๊ฒ ๋‹ค๊ณ  ํ•จ. ์˜ค์ฃฝํ•˜๋ฉด ์ €์ปค๋ฒ„๊ทธ๊ฐ€ "์ง€๊ธˆ ํ…Œํฌ ์—…๊ณ„์—์„œ ์šฐ๋ฆฌ ํšŒ์‚ฌ์— ์ปดํ“จํŒ… ์ž์›์ด ๋‚จ์•„๋ˆ๋‹ค๊ณ  ๋А๋ผ๋Š” ์‚ฌ๋žŒ์€ ๋‹จ ํ•œ ๋ช…๋„ ์—†๋‹ค"๊ณ  ๋ชป์„ ๋ฐ•์•˜๊ฒ ์–ด. ์—ฌ๊ธฐ์„œ ๋‚ด๊ฐ€ ์ง„์งœ ๊ฐ•์กฐํ•˜๊ณ  ์‹ถ์€ ์ƒ๊ฐ์€, ์ด๊ฑด 90๋…„๋Œ€ ๋‹ท์ปด ๋ฒ„๋ธ”์ด๋ž‘ ๋ณธ์งˆ์ ์œผ๋กœ ์™„์ „ํžˆ ๋‹ค๋ฅด๋‹ค๋Š” ๊ฑฐ์•ผ. ๊ทธ๋•Œ๋Š” ์‹ค์ฒด ์—†๋Š” ๊ธฐ๋Œ€๊ฐ๋ฟ์ด์—ˆ๋‹ค๋ฉด, ์ง€๊ธˆ์€ '์••๋„์ ์ธ ์ธํ”„๋ผ ๊ทœ๋ชจ์˜ ๊ฒฝ์ œ'์™€ '์ž์ฒด ์‹ค๋ฆฌ์ฝ˜ ์นฉ ์—ญ๋Ÿ‰'์ด๋ผ๋Š” ๊นกํŒจ ๊ฐ™์€ ๋ฌผ๋ฆฌ์  ์ž์‚ฐ์ด ์‹ค์‹œ๊ฐ„์œผ๋กœ ๋ˆ์„ ์ฐ์–ด๋‚ด๊ณ  ์žˆ์Œ. ๋ฉ”ํƒ€๊ฐ€ ์„ฑ๋Šฅ ์ข‹์€ API๋ฅผ ํƒ€์‚ฌ ๋Œ€๋น„ ํŒŒ๊ฒฉ์ ์œผ๋กœ ์‹ผ ๋‹จ๊ฐ€์— ๋ฟŒ๋ฆด ์ˆ˜ ์žˆ๋Š” ๋ฌด๊ธฐ๋„ ๊ฒฐ๊ตญ ์ด ๊ฑฐ๋Œ€ํ•œ ์ธํ”„๋ผ๊ฐ€ ๋ฐฑ์—…ํ•ด์ฃผ๋‹ˆ๊นŒ ๊ฐ€๋Šฅํ•œ ์ผ์ด์ง€. ๊ฒฐ๊ตญ ํ˜„์‹œ๋Œ€์˜ ์›์œ ๋Š” 'Compute(์—ฐ์‚ฐ ์ž์›)'๊ณ , ์ด๊ฑธ ์–ผ๋งˆ๋‚˜ ์ฅ๊ณ  ์žˆ๋А๋ƒ๊ฐ€ ๋ฏธ๋ž˜ ํด๋ผ์šฐ๋“œ ์‹œ์žฅ ์ฃผ๋„๊ถŒ์ด๋ž‘ ์ง๊ฒฐ๋ผ. ๋น…ํ…Œํฌ๋“ค ์ž…์žฅ์—์„  ์ธํ”„๋ผ๊ฐ€ ๊ณง ๋ˆ์ธ๋ฐ ํˆฌ์ž๋ฅผ ์ค„์ผ ์ด์œ ๊ฐ€ ์ „ํ˜€ ์—†์ง€. ์˜คํžˆ๋ ค ๋‚จ๋“ค๋ณด๋‹ค ๋” ๊ฑฐ๋Œ€ํ•˜๊ฒŒ ์ง€์œผ๋ ค๊ณ  CapEx ์ „์Ÿ์— ๊ธฐ๋ฆ„์„ ๋ถ€์„ ๋ช…๋ถ„์ด ์ƒ๊ธด ๊ฑฐ์•ผ. AI๋Š” ๋๋ฌผ์ด ์•„๋‹ˆ๋ผ ์ด์ œ ๋ง‰ ์ง„์งœ ๋น„์ฆˆ๋‹ˆ์Šค๋กœ ์ง„ํ™”ํ•˜๋Š” '์‹œ์ž‘ ๋‹จ๊ณ„'์ž„. ์ด ์ธํ”„๋ผ ํŒจ๊ถŒ ์ฅ๋Š” ๋†ˆ์ด ๋‹ค ๋จน๋Š”๋‹ค ์ง„์งœ.
ํ•œ๊ตญ์–ด
1
3
29
3.3K