The MeeCrypt

Fluence has launched a competitive auction marketplace for GPU capacity at http://auctions.fluence.ai, where demand and supply for decentralized compute can meet directly.

Businesses and AI teams post their compute requirements, including specs and rental window. Providers with available capacity then place competitive bids. The best bid wins, and all terms including price, SLA, rental duration, and bid history are recorded on-chain for transparency.

The old model forced teams to over-provision. Buy hardware for peak demand and accept low utilization the rest of the time. The new model allows teams to reserve capacity only when needed and let price discovery happen through open bidding.

This is significant for three reasons:

1. Cost efficiency: Pay for compute when needed, not for idle hardware.
2. Monetization: Providers can turn underutilized GPUs into active revenue.
3. Transparency: Auction history, pricing and SLAs are visible, creating a real market for compute.

The marketplace is live. Post a request or list capacity at http://auctions.fluence.ai

#Fluence #GPU #Compute

GPU Ownership vs Renting

One thing a lot of teams get wrong with AI infrastructure is focusing on the price of the GPU, not how much they'll actually use it.

Buying hardware feels like the right move when you need power. But the better question is: how many hours a day is that hardware actually working?

A server running 20+ hours a day is a completely different investment from one that sits idle 4 days a week.

If your utilization is consistently high, owning makes sense. You spread the upfront cost, power, cooling, maintenance across a ton of productive hours.

The problem is when workloads are unpredictable.

Most teams have peaks and valleys. Heavy inference, training sprints, dev work some days, then almost nothing the next. But the costs don't stop when usage drops. You still pay for the hardware, power, cooling, maintenance, and replacement.

That's where renting compute changes the math.

Instead of buying for peak demand, you rent capacity when you actually need it.

@Fluence gives access to decentralized GPU infrastructure. So you don't have to own and maintain a full fleet just to cover the spikes.

When demand jumps, you can scale up with rented compute instead of making a big capital bet on hardware that might be underutilized next week.

It shifts how you think about infra spend:

Ownership works when utilization justifies it. Renting works when flexibility is the priority.

Renting isn't always cheaper pricing, availability, latency, data needs all matter. And GPUs evolve fast. What is top tier today won't be the most efficient in 2-3 years. Renting helps avoid getting stuck with aging hardware.

For me, the interesting part about Fluence isn't just GPU access. It's treating compute as something you tap into when needed, not an asset you have to justify every day.

Because having powerful GPUs is one thing. Making sure they actually earn their keep is another.

#Fluence #DePIN #GPU

For a long time, decentralized compute sounded good in theory but was hard to use in practice.

For developers and teams building products, the flexibility @Fluence offers is a game-changer. You get to build without worrying about the hidden costs and restrictions that come with traditional centralized cloud providers.

Here’s what theyve got👇

1. Zero Egress Fees: Move your data in and out without being charged extra. What you use is what you pay for.

2. Unlimited Bandwidth: No throttling, no surprise bills when your app scales.

3. Open Standards: Built on open protocols, not closed systems. You own your stack.

4. Easy Migration & No Vendor Lock-in: Deploy where you want, when you want. Migrate workloads in minutes, not months.

Training models, running inference, storing massive datasets, it all needs scalable, affordable, and flexible infrastructure. Having control over WHERE your workloads run and HOW you scale is no longer a luxury, it's a necessity for any serious builder.

Fluence is giving developers that freedom. The freedom to choose, to scale, and to build without limits.

Builders are already making the switch.

#Fluence #DePIN #DecentralizedCloud #AI

@Fluence Updates Its Compute Network With Two New NVIDIA B300 Clusters

Two NVIDIA HGX B300 clusters are now open for reservation through Fluence, bringing more high-performance GPU capacity to its compute network.

Here’s the interesting part:

🇺🇸 Texas: 10 liquid-cooled nodes, with 8 B300 GPUs per node. Each GPU has 288 GB of HBM3E memory, giving each node 2.3 TB of GPU memory. The cluster can be ready within one week of signing.

🇺🇸 Nevada: 6 air-cooled nodes using Aivres KR9288X3 systems with HGX B300 GPUs, expected to be ready on January 1, 2027.

Both options offer dedicated bare-metal infrastructure, full root access, 2 TB of system memory, and 30.72 TB of NVMe storage.

What I find interesting here is the practical side of the update. These aren't just new hardware specifications to read about. Teams can reserve the infrastructure for their own workloads and have direct control over their systems.

As AI workloads become more demanding, I think the infrastructure powering them deserves just as much attention as the applications everyone sees.

Fluence is continuing to build in that space, and I'm curious to see how its compute network develops from here.

#Fluence #AI #DePIN

Fluence bilan bulut xarajatlarini 3 qadamda kamaytirish:

1-qadam: Fluence konsoliga kiring
AWS’ga kirmang. http://fluence.ai/gpu sahifasiga o‘ting.
Hamyoningizni ulang — boshqaruv paneli ochiladi.

2-qadam: Bulut xizmatini emas, GPU’ni ijaraga oling
«GPU ijaraga olish» tugmasini bosing.
Keraklisini tanlang: H100, A100 yoki H200. O‘zingizga yaqin joylashuvni tanlang.
Fluence’da 32 ta hududda 1400 dan ortiq GPU bor. Narx AWS/GCP’dan allaqachon 60% arzon. Soatlik narxni real vaqt rejimida ko‘rasiz.

3-qadam: Faqat foydalanganda to‘lang
GPU’ni ishga tushiring, modelingizni joylashtiring va uni o‘qiting.
Ishingiz tugagach, UNI O‘CHIRING.
Fluence har soniya uchun hisob-kitob qiladi. Oylik majburiyat yo‘q. 2 soat foydalansangiz, faqat 2 soat uchun to‘laysiz.

Shu, xolos.

Kreditlar yo‘q, savdo bo‘yicha qo‘ng‘iroq yo‘q, korporativ shartnoma yo‘q.

An’anaviy bulut xizmati: soatiga $3.50 to‘laysiz, ishlatilmay turgan paytda ham to‘lov davom etadi — oyiga $2500.
Fluence: soatiga taxminan $1.40 to‘laysiz va faqat faol ishlatilgan vaqt uchun to‘laysiz — oyiga taxminan $900.

Xarajatlarni 60% kamaytirasiz, chunki:
1. Har soat uchun kamroq to‘laysiz
2. Faqat faol foydalanilgan vaqt uchun to‘laysiz

#Fluence #GPU #DePIN #AI

@Fluence DAO’ning 2026-yil 3-chorak hisoboti e’lon qilindi va men $FLT’ga yanada optimistik qaray boshladim

Hisobotning qisqacha mazmuni:

1. Ular juda kam mablag‘ sarflab, ko‘p ish qilishyapti.
3 oyda staking mukofotlarini hisobga olmaganda, atigi $300 ming sarflashdi. To‘liq bulutli platforma yaratayotgan texnologiya kompaniyasi uchun bu juda kam. Aksariyat kompaniyalar bundan 10 baravar ko‘p sarflaydi.

Shunga qaramay, ularning xazinasida hali ham 350 milliondan ortiq FLT va 1,1 million dollardan ortiq steyblkoin bor. Demak, ular uzoq vaqt davomida rivojlanishda davom eta oladi.

2. Odamlar foydalana oladigan haqiqiy mahsulotlarni ishga tushirishyapti.
• Virtual Private Cloud’ni ishga tushirishdi — bu kompaniyalarga ilovalarini yanada xavfsiz va maxfiy tarzda ishlatish imkonini beradi.
• Oyiga $2.54 dan boshlanadigan arzon kompyuterlarni ishga tushirishdi, shuning uchun hatto kichik dasturchilar ham ulardan foydalana oladi.
• Web Terminal’ni qo‘shishdi, endi undan to‘g‘ridan-to‘g‘ri brauzeringiz orqali foydalanishingiz mumkin.
• AI agentlari USDC bilan mustaqil to‘lov qilib, kompyuterlarni avtomatik ijaraga olishi mumkin bo‘lgan funksiyani qo‘shishdi. Ro‘yxatdan o‘tish uchun odam kerak emas.

Eng katta yangilik esa — GPU auksionlari. Ular kompaniyalarga AI uchun kuchli GPU’larni ijaraga olishda yordam beradi. Faqat 3-chorakning o‘zida ular qiymati $6M, $15.4M va $5.5M bo‘lgan bitimlarning tuzilishiga yordam berishdi, jumladan, Osiyodagi 96 ta H100 bo‘yicha kelishuv ham bor. Bu haqiqiy biznes.

Shuningdek, ular http://fluence.ai saytiga o‘tishdi — bu endi AI bulutiga e’tibor qaratayotganlarini ko‘rsatadi.

3. Token egalari daromad olishda davom etyapti.
3-chorakda staking uchun 990 ming FLT to‘lashdi, yillik foiz stavkasi taxminan 11.5%. Iyulda 340 ming, avgustda 300 ming, sentabrda 350 ming.

Men uchun hammasi oddiy: kam xarajat, yaxshiroq xazina, yaxshi mahsulotlar va GPU’lar uchun millionlab to‘layotgan haqiqiy mijozlar.

FLT o‘sishini kuzatishda davom etaman

#FLT #Web3 #DePIN

@Fluence yana 2,4 mln dollar jalb qildi, endi worldengine_ai bilan Jismoniy AI’ni quvvatlaydi

Fluence uchun yana bir kun, yana bir haqiqiy shartnoma. Bu safargisi esa hikoya uchun juda muhim.

Fluence hozirgina WorldEngine AI bilan Jismoniy AI uchun inferens hisoblash quvvatini taqdim etish bo‘yicha 2,4 mln dollarlik kelishuv imzoladi.

Mana nima uchun bu $FLT egalari uchun juda muhim deb o‘ylayman:

1. Haqiqiy daromad, haqiqiy foydalanish holati
Bu grant yoki testnet hamkorligi emas. Bu amaldagi inferens vazifalari uchun 2,4 mln dollar. DePIN’da bunday kelishuvlar ko‘proq bo‘lishi kerak.

2. Jismoniy AI > Raqamli AI
Hozir har kim modelni o‘qita oladi. To‘siq esa uni ishga tushirishda. Robotlarga 24/7 rejimida past kechikishli, ishonchli hisoblash quvvati kerak. Inferens ishlamay qolsa, robot ham ishlamay qoladi. Fluence aynan shu muammoni hal qilmoqda.

3. Texnologik stek bilan uyg‘unlik mukammal
WorldEngine quyidagi yo‘nalishlarda ishlamoqda:
→ robototexnika ma’lumotlari
→ modelni o‘qitish
→ baholash
→ ishga tushirish

Fluence yakuniy va eng muhim qatlam — inferensni quvvatlaydigan markazlashmagan hisoblash infratuzilmasi sifatida ayni mos keladi.

Men bir necha oydan beri aytib kelaman: AI’ning kelajagi jismoniy olamda. AI jismoniy dunyoga kirib kelganda, g‘oliblar faqat model yaratuvchilar bo‘lmaydi. Hammasini ishlatib turadigan infratuzilma ta’minotchilari ham g‘olib bo‘ladi.

Bunday hamkorliklar qancha ko‘p bo‘lsa, $FLT’ga talab ham shuncha ortadi.

AI tanaga ega bo‘lmoqda, Fluence esa uning miyasini quvvatlayapti.

«Ijobiy kayfiyatdaman» deyish buning oldida kamlik qiladi.

#FLT #DePIN #Web3 #AI

A lot of infrastructure products make developers leave their existing workflow to use them.

@Fluence MCP is moving in the opposite direction.

Once connected, your AI coding agent can work with Fluence compute directly from the environment you already use.

The setup is lightweight:

• One config snippet
• A Fluence API key
• An MCP-compatible client

No local MCP server to install. No updates to manage.

And after the initial connection, you don’t need to keep opening the Fluence Console. The interaction stays inside your coding client.

That may sound like a small UX improvement, but I think it matters.

Developers are already getting comfortable with AI agents handling more of the development process. Giving those agents access to compute is a logical next step.

#FluenceProject #DePIN #AI

DePIN still struggles with is making decentralized infrastructure feel accessible to developers.

@Fluence is taking an interesting approach with its new MCP integration. Which means AI agents can interact with Fluence compute through MCP-compatible tools.

• Connect Cursor, Claude Code, VS Code, Codex or another MCP client
• Tell your AI agent what compute you need
• Let the agent handle the interaction with Fluence
• Get decentralized compute without making infrastructure management the center of your workflow

Developers don’t necessarily want another complicated infrastructure stack to learn. They want to build, deploy and let the underlying systems do their job.

If Fluence can make decentralized compute something an AI agent can simply call when needed, that gives DePIN a much more practical developer story.

The takeaway for builders is simple: don’t just watch the DePIN narrative. Try the infrastructure.

I genuinely hope more developers discover Fluence and see what they can build with it.

#FLT #DePIN #Web3 #AI

Men hamyonimda @Fluence in dan FLT borligini deyarli unutib qo'ydim.

O'tgan yili bir nechta FLT olib qo'ygandim va rostini aytsam, biroz vaqt bunga unchalik e'tibor bermagandim.

Lekin yaqinda Fluence’ni yana bir bor ko'rib, token aslida ekotizim ichida nima qilishini tekshirib chiqishga majbur bo'ldim.

Va o'ylaganimdan ham ko'proq narsa bor ekan.

• FLT staking tizimining bir qismi: tokenlar hisoblash resurslarining xavfsizligi va koordinatsiyasi atrofida ishlatiladi.

• Uning governance (boshqaruv)da ham roli bor — jamoaga Fluence ekotizimi bo‘yicha qarorlar qabul qilishda ishtirok etish imkonini beradi.

• Hisoblash provayderlari va tarmoq iqtisodiyoti FLT bilan bog‘langan — bu tokenning faqat hamyon ichida yotib qolishdan tashqari amaliy vazifasi borligini ko‘rsatadi.

• Bulut tomoni ham tobora qiziqroq bo‘lyapti. Fluence foydalanuvchilarga an’anaviy markazlashgan bulut provayderlariga butunlay tayanmasdan hisoblash resurslariga kirish imkonini beradigan infratuzilma qurmoqda.

• Hozir esa yanada kattaroq yo‘nalish bor: AI va GPU hisoblash. AI ilovalari talab qilayotgan hisoblash quvvati miqdori bilan GPU’ga markazlashmagan tarzda kirish tobora kuzatishga arziydigan yo‘nalishga aylanmoqda.

Oxirgi qismi, ehtimol, meni eng ko‘p o‘ziga tortgan.

Chunki o‘tgan yili o‘sha bir nechta FLT’ni olganimda, AI infratuzilmasi yoki GPU talabi haqida unchalik o‘ylamagan edim.

Endi esa Fluence atrofidagi suhbat butunlay boshqacha ko‘rinadi.

Men FLT keyin qayerga borishini bilaman demayman.

Shunchaki, siz bexosdan olib qo‘ygan token asta-sekin ancha katta infratuzilma hikoyasining bir qismiga aylanib borayotgani qiziq tuyuladi.

Ha, demak… men yana kichkina FLT paketimni kuzatib yuraman.

#Fluence #Web3 #DePIN #AI

@Fluence has officially moved its website from Fluence Network to fluence.ai, and I think the change says more about the project’s direction than it might seem at first.

• New domain: Fluence Network is now represented through fluence.ai, putting AI much closer to the center of the project’s identity.

• AI infrastructure: This reflects Fluence’s growing focus on infrastructure for AI workloads, rather than simply being known as a decentralized compute network.

• GPU + CPU compute: Fluence is building around access to both GPU and CPU resources, which are becoming increasingly important as AI workloads continue to grow.

• Fluence Cloud: One interesting detail is that Fluence now describes Fluence Cloud as the successor to the Fluence Project.

• A broader infrastructure focus: The project’s direction appears to be moving toward cloud infrastructure and making distributed compute more accessible to developers and builders.

• The domain tells part of the story: Moving from “Network” to “.ai” gives the brand a more direct connection to the AI infrastructure space it is increasingly focused on.

It’s still just a domain change on the surface, but when you look at the bigger picture, it feels like a clearer signal of where Fluence is heading.

From decentralized compute to AI-focused cloud infrastructure, the evolution is becoming easier to see.

#Fluence #AI #DePIN #Web3

Your cloud bill might be more flexible than you think.

Cloud infrastructure is convenient, but convenience can get expensive when compute demand starts scaling especially for AI workloads.

@Fluence $FLT takes a different approach by connecting workloads to a decentralized network of compute providers.

• Lower infrastructure costs: access to distributed compute without relying entirely on traditional cloud providers
• More flexibility: workloads can tap into compute from different providers
• AI-ready infrastructure useful as AI applications continue increasing demand for compute
• Less cloud concentration: an alternative to putting everything through the same major providers

The bigger idea is simple: compute is becoming increasingly important, so having more options for where and how it is sourced could matter just as much as having more of it.

That’s where decentralized compute gets interesting.

#Fluence #DePIN #AI

Everyone said $ETH . I started thinking about $FLT from @Fluence

I saw an interesting question recently: if $BTC suddenly disappeared, which asset could take its place?

The answers were mostly ETH and other established crypto names.

I started thinking about a different angle: what if the next important narrative isn't just another financial asset, but the infrastructure powering the next wave of technology?

AI is already creating massive demand for compute. Models, agents, and applications all need computing resources to operate.

That’s why $FLT has been interesting to watch.

@Fluence is working toward decentralized compute, and as Evgeny Ponomarev discussed in a recent AMA, the bigger vision is to create an alternative to increasingly centralized cloud infrastructure, with AI being an important starting point.

$BTC represents a major chapter in decentralized finance.

Could decentralized compute become an equally important piece of the next chapter of Web3?

@Fluence : Kimga GPU quvvati beriladi?

GPU quvvati qanday qilib aslida sotib olinadi va sotiladi. Dunyoda juda ko‘p GPU bo‘lishi mumkin, lekin baribir muayyan xaridor, joylashuv, ish yuklamasi yoki vaqt oralig‘i uchun tanqislik bo‘lib qoladi. Shuning uchun narxlash va kirish imkoniyati GPU yoniga oddiy raqam qo‘yishdan ko‘ra murakkabroq.

Aynan shu yerda auktsionlar foydali bo‘ladi. Oldinga yo‘nalgan auktsionda xaridorlar mavjud quvvat uchun raqobatlashadi, teskari auktsionda esa provayderlar talabni qondirish uchun raqobatlashadi.

Ikkalasi ham berilgan lahzada hisoblash resursining haqiqiy qiymatini aniqlashning yanada dinamik usulini yaratadi. Jiddiy GPU quvvatini sotib olayotgan odam uchun esa eng arzon variant avtomatik tarzda to‘g‘ri tanlov bo‘lib qolmaydi.

Yana shuni so‘rash kerak: Quvvat sizga kerak bo‘lgan paytda haqiqatan ham mavjudmi? Tarmoq (networking) matosi ish yuklamasini ko‘tara oladimi? Infrastruktura bilan qanday SLA keladi? Agar nimadir buzilib qolsa, u qanchalik tez tiklanishi mumkin? Bu tafsilotlar AI ish yuklamalari ko‘lamda ishlaganda katta farq qilishi mumkin.

Fluence rezerv qilingan GPU klasterlari uchun malakali talab bilan ma’lumotlar markazlarini bevosita bog‘laydigan global bozor tomon rivojlanmoqda.

Qiziqarli tomoni shundaki, gap faqat ko‘proq GPU qo‘shish haqida emas. Bu allaqachon mavjud bo‘lgan GPU’larni topish, taqqoslash, narxlash va foydalanishni osonlashtirish haqida.

#Fluence #DePIN #AI

Orqa fonda AI’ni kim quvvatlayapti? @Fluence

Menga AI agentlari haqida qiziq tuyuladigan bir narsa shuki, kechikish (latency) hamisha modelning o‘zidan kelmaydi.

Agent bir nechta tool chaqirishi, holatni (state) olishi, tekshiruvlar (validations)ni bajarishi, xotirani (memory) yangilashi va bir necha marta takrorlanadigan siklni davom ettirishi mumkin. Har bir bosqichda bir necha millisekundlik CPU kechikishi o‘z-o‘zicha unchalik katta ko‘rinmasligi mumkin, ammo butun ish jarayoni bo‘ylab bu kechikishlar yig‘ilib, so‘rovni uning belgilangan muddatidan (deadline) o‘tkazib yuborishi mumkin.

Shu yerda infratuzilma dastlab ko‘ringanidan ham muhimroq bo‘lib qoladi.

AI yuklamalari uchun o‘rtacha (average) ko‘rsatkich har doim ham yetarli emas. Asosiy muammo, ayniqsa CPU resurslari uchun raqobat (contention) ko‘plab bosqichlar bo‘ylab kichik kechikishlarga sabab bo‘lganda, yuzaga chiqadi.

Shuning uchun oldindan taxmin qilinadigan (predictable) ishlash muhim bo‘lgan yuklamalar uchun ajratilgan hisoblash resurslari (dedicated compute) qiziqarli variantga aylanadi.

Fluence buni markazlashmagan (decentralized) hisoblash tomondan yondashib, ishlab chiquvchilarga ko‘proq barqaror sig‘imga (capacity) ehtiyoj bo‘lgan yuklamalar uchun ajratilgan CPU resurslari imkonini bermoqda.

AI agentlari oddiy tajribalardan ishlab chiqarish (production) bosqichiga o‘tayotgani sari bu tobora dolzarb bo‘lib borayotganini o‘ylayman. Agent o‘nlab bosqichlarni bajarayotganda, har bir bosqichdagi o‘zgaruvchanlikni (variability) kamaytirish yakuniy foydalanuvchi tajribasiga ancha katta ta’sir ko‘rsatadi.

Menga kuzatayotgan savol oddiy: agent yuklamalari murakkablashgani sari, oldindan taxmin qilinadigan hisoblash resurslari faqat xom (raw) hisoblash quvvati kabi muhim bo‘lib qoladimi?

https://fluence.ai/

#Fluence #AI #DePIN

@Fluence VPC ishga tushishi Fluence ekotizimi uchun muhim qadam bo‘lishi mumkin.

https://fluence.ai/

Yangi VPC infratuzilmasi ishlab chiquvchilarga xususiy tarmoq bilan bir nechta VM’larni ishga tushirish, security guruhlarini qo‘llash, nazorat qilinadigan egress, ixtiyoriy public IP’lar va cloud-init’ni taqdim etadi. Oddiy qilib aytganda, bu Fluence’ni turli xizmatlar birgalikda xavfsiz ishlaydigan real ilovalarni ishga tushirish uchun yanada mosroq qiladi.

$FLT nuqtayi nazaridan, men uchun bu yerda qiziqarli jihatlar boshlanadi.

Yangi infratuzilma funksiyasi avtomatik ravishda FLT narxi oshishini anglatmaydi. Muhimroq savol shuki, VPC Fluence’ga ko‘proq ishlab chiquvchilar, yuklamalar va bizneslarni jalb qilishga yordam beradimi?

Agar qabul (adoption) o‘sib borsa, tarmoq bo‘ylab ko‘proq kompyut resurslari joylashtirilishi mumkin, bu esa kuchliroq ichki talabni yaratib, ehtimol FLT’ni qo‘llab-quvvatlaydigan asoslarni yaxshilaydi.

Shuning uchun men e’lonning o‘zidan ko‘ra undan keyingi foydalanishni kuzataman. Agar Fluence amaliy infratuzilma qo‘shishni davom ettirsa va bu funksiyalar real tarmoq faolligiga aylansa, vaqt o‘tishi bilan bu FLT uchun ijobiy katalizator bo‘lishi mumkin, deb o‘ylayman.

Men uchun tezis oddiy: yaxshiroq infratuzilma → ko‘proq foydalanish holatlari → ko‘proq tarmoq faolligi → ehtimol kuchliroq FLT asoslari.

#DePIN #AI #Web3

@Fluence uchun yana bir $5.5M hisoblash bitimi

Fluence biologiyadan dorilarga tezroq yo'llarni topishga qaratilgan AGI yaratib borayotgan mijoz bilan $5.5M qiymatidagi hisoblash bitimini ta'minladi.

Menga ko'zga tashlangani faqat bitimning hajmi emas.

Hal qilinayotgan muammo muhim: hisoblash quvvati tor bo'g'inga aylanib qolmasligini ta'minlash.

AI rivoji tobora ko'proq hisoblash resurslarini talab qilmoqda, lekin shunchaki ko'proq GPU'ga ega bo'lishning o'zi yetarli emas. Asosiy qiyinchilik — hisoblash quvvatini kerak bo'lgan joyda va kerak bo'lgan paytda mavjud qilish.

Mana shu yerda markazlashmagan GPU bozorlari qiziqarli bo'ladi.

Fluence GPU ta'minotchilari ish yuklari uchun raqobatlasha oladigan maydoncha qurmoqda, bu esa jamoalarga kerakli hisoblash quvvatiga yanada moslashuvchan usulda kirish imkonini yaratishi mumkin.

Menimcha, AI'ning o'zi emas, balki uning ortidagi hisoblash qatlami haqida ham ko'proq eshitamiz.

$5.5M bu bitta bitim, ammo u ishonchli GPU hisoblash quvvatiga kirish qanchalik qimmatga aylanib borayotganini ham ko'rsatadi.

Fluence keyin nima qilishini kuzatishga arziydi.

$FLT #DePIN #AI #Web3

@Fluence : Yashirin GPU bozori

AI ish yuklamalari kengaymoqda, data-markazlar kattalashmoqda va kompaniyalar hisoblash quvvatiga katta mablag‘ sarflamoqda.

Ammo GPU hikoyasining menimcha e’tiborga arzigulik yana bir tomoni bor:

Ba’zi uskunalar ish yuklamalari orasida bekor turadi. Ba’zilari to‘liq ishlatilmaydi. Yana ba’zi hisoblash resurslariga boshqalar uchun shunchaki kirish qiyin.

Bu esa kelajakdagi hisoblashni shunchaki ko‘proq GPU qo‘shishdan ko‘ra, mavjud GPU’lardan qanchalik samarali foydalana olishimizga bog‘liqroq qiladi.

@Fluence g‘oyasi sodda, ammo qiziqarli: mavjud hisoblash quvvatini unga ehtiyoji bor odamlar va ish yuklamalari bilan bog‘lay oladigan infratuzilma yaratish.

Agar AI talabi tezlashishda davom etsa, menimcha, muhokama asta-sekin quyidagilardan:

“Kimda eng ko‘p GPU bor?”

quyidagiga o‘tadi:

“Kim eng ko‘p hisoblash quvvatini mavjud, moslashuvchan va samarali foydalaniladigan holatda taqdim eta oladi?”

GPU iqtisodi endi faqat uskuna haqida bo‘lmasligi mumkin.

U mavjud hisoblash quvvati atrofida yaxshiroq bozorlar yaratish haqida ham bo‘lishi mumkin.

#DePIN #AI #Web3

Qog‘ozdagi eng tezkor GPU har doim ham har bir ish yuklamasi uchun eng yaxshi tanlov degani emas.

Benchmark sizga qaysi GPU soniyada eng ko‘p tokenni chiqarishi mumkinligini ko‘rsatishi ehtimol, ammo real joylashtirishlarda hisobga olinishi kerak bo‘lgan ko‘proq omillar bor.

Model hajmi, chiqish uzunligi, xotira talablari, parallellik (concurrency), mintaqa (region), mavjudlik (availability) va, albatta, ish yuklamasini ishga tushirishning amaldagi narxi hammasi tenglamani o‘zgartirishi mumkin.

Shunda @Fluence qiziqarli yondashuvni taklif qiladi.

GPU Cluster Auctions orqali xaridorlar turli provayderlarning takliflarini GPU turi, quvvati (capacity), mintaqasi, narxlari va mavjudligi bo‘yicha solishtirishi mumkin.

Bu qarorni faqat eng yuqori unumli (highest-performing) apparatni quvishdan uzoqlashtiradi.

Shunchaki so‘rash o‘rniga:

“Qaysi GPU eng tez?”

Yaxshiroq savol bo‘lishi mumkin:

“Qaysi GPU menga kerak bo‘lgan unumdorlikni, aslida mantiqiy narxda beradi?”

Chunki yanada kuchliroq GPU sezilarli darajada qimmatroq bo‘lsa-yu, sizning aniq ish yuklamangiz uchun muhim ustunlik bermasa, qo‘shimcha unum uchun pul to‘lash o‘zini oqlamasligi mumkin.

Agar boshqa GPU o‘sha ish yuklamasini arzonroq narxda bajara olsa, bu farq miqyosda ayniqsa ahamiyat kasb eta boshlaydi.

AI bo‘yicha ish yuklamalari rivojlanishda davom etar ekan, GPUga talab faqat ko‘proq hisoblash quvvatini topish bilan bog‘liq bo‘lmaydi.

Bu, shuningdek, to‘g‘ri joyda, to‘g‘ri narxda, to‘g‘ri hisoblash (compute)ni topish haqida bo‘ladi.

Shuning uchun GPU narxini aniqlash (price discovery) tobora muhimlashib borishi mumkin.

#Flt #DePIN #Web3 #AI

@Fluence GPU KIMOSH (AUCTIONS) YANA TUSHDI: YANA 15,4 Mln $+ LOKALASHGAN BITIM

Fluence o‘z auktsion platformasi orqali yana bir yirik GPU bitimini qo‘lga kiritdi.

Bu bitim Osiyodagi 96 ta H100 tugunni qamrab oladi, AI inference va model trenirovkasi uchun ishlab chiqarishga o‘tmoqda.

Eng qiziq tomoni bitim hajmida emas.
Bu raqobatli savdolar orqali real GPU talabi harakatlanayotganini ko‘rsatadi: imkoniyat (sig‘im) izlayotgan xaridorlarni GPU’lari tayyor bo‘lgan provayderlar bilan bog‘laydi.

Platforma orqali yanada yuqori qiymatli bitimlar kelib tushishi bilan GPU Auctions AI infratuzilmasi bozori kuzatishga arziydigan qismiga aylanib bormoqda.

#FLT #DePIN #Web3