Points of inference in AI

“NVIDIA, A MANUFACTURER of computer chips, is the most valuable company in the world. It owes its success to the versatility of the graphics processing unit (GPU), a chip it pioneered in the late 1990s.

 

Originally designed to make video games look better, GPUs turned out to be well suited to training large language models (LLMs).

 

That discovery sent demand for Nvidia’s chips, and its valuation, soaring.

 

Times are changing fast.

 

Demand for AI computing is shifting from training models to getting them to answer real-world queries, a process known as inference.

 

McKinsey, a consultancy, estimates that by the end of the decade inference will account for three-fifths of demand in AI data centres.

 

Nvidia appears to recognise the shift. On March 16th it unveiled a new chip designed specifically for inference tasks, the Groq 3 LPX, with an architecture that departs from the traditional GPU.

 

This time, it will have plenty of competition. A crop of startups is building chips aimed at running AI models faster and more efficiently than Nvidia’s.

 

Training and inference place different demands on hardware.

 

 Training, in which an AI model is taught to identify patterns in vast amounts of raw data, relies on enormous numbers of calculations being conducted in parallel.

 

Nvidia’s B200 chip, for instance, one of the company’s flagship products, contains more than 16,000 processing units, also known as cores, to perform such operations.

 

Inference, in which a finished model calls on its training to respond to user prompts, works differently. It unfolds in two stages: prefill and decode.

 

During prefill, the model processes the prompt and converts it into small units of text, typically about four characters in English, known as tokens. To speed things up, tokenising different parts of the query can be done in parallel. Decoding then generates the response, token by token. To do this, the model relies on its “weights” (relationships between tokens learned during training) as well as previously generated tokens. These weights are stored in the system’s memory.

 

The need for constant memory access is where modern GPUs fall down. AI processors like the B200 contain small but extremely fast on-chip memory, known as SRAM, as well as a much larger off-chip memory known as DRAM. Accessing DRAM can be ten times slower and consume far more energy than reading SRAM. The problem is worsening. As AI models grow larger and become better at handling long user prompts, their memory demands are rising sharply. A study by Amir Gholami of the University of California, Berkeley, and colleagues finds that over the past two decades computing performance has roughly tripled every few years, whereas off-chip memory bandwidth has improved by a factor of only about 1.6. This “memory wall” has become the main bottleneck in increasing the speed of AI inference.

 

You must remember this

 

GPUs rely on software workarounds to cope. One approach splits the two stages across different processors. The prefill phase runs on GPUs optimised for high parallel computing power, while decoding runs on separate GPUs designed for fast memory access. Another technique is batching, where many queries are processed together. Once the model’s weights are loaded, they can then be used for many queries at the same time, reducing repeated trips to the external memory.

 

Nvidia’s new chip uses the power of software to give the on-chip memory a boost. The size of the SRAM is around 500 megabytes—tiny when compared with the B200’s 192 gigabytes of off-chip memory. What makes the difference is smart software that choreographs how every piece of data moves through the chip to maximise computation and memory access.

 

Startups are experimenting with more radical designs. One approach is to simply build a bigger chip. That is the approach taken by Cerebras, an American chip designer. Its latest chip, the size of a dinner plate, contains an enormous 900,000 cores and 44 gigabytes of on-chip SRAM. Because all data movement occurs within the wafer, Cerebras claims its system can run inference up to 15 times faster than conventional designs. For very large models, however, storing all their parameters on SRAM is impractical.

 

Others are tackling the problem by redesigning how data move through the cores. MatX, a startup founded by former Google chip engineers, builds on an idea used in Google’s tensor processing units (TPUs). These chips rely on what is called a systolic array, a grid of processing elements through which data flow rhythmically, rather like blood pumped through the body. After each calculation the result passes directly to the next unit, bypassing the need to store intermediate results in memory. Traditional systolic arrays, however, are fixed in size. Make them bigger, for larger tasks, and they will often sit idle; make them smaller, and efficiency falls when the larger tasks come through. MatX proposes a “splittable” systolic array that divides the processor into several smaller grids, allocating computing resources differently depending on whether the chip is handling prefill or decode.

 

A third approach, pursued by d-Matrix, a California-based startup, tries to eliminate the memory wall entirely by having the same components handle both memory and computation. This architecture, known as in-memory computing, promises lower energy use and faster inference.

 

Others advocate chip designs built around specific algorithms to improve efficiency further. Etched, another Californian startup, is designing a chip custom-built to run transformer models, the algorithms that underpin most LLMs. This specialisation allows the company to strip away hardware needed for other uses and simplifies the software running on the chip.

 

Researchers in China have proposed an even more radical form of specialisation: embedding model weights directly into hardware. In one design from the Chinese Academy of Sciences, these are physically encoded in the layout of metal wires. The authors claim this technique removes the need to fetch parameters from memory, enabling extreme efficiency.

 

Yet such specialisation carries risks. Designing a new chip typically takes 12–18 months, whereas AI algorithms evolve far faster. A chip built around today’s dominant model architecture could quickly become obsolete if the field shifts.

 

The chips have yet to fall. Nvidia’s rivals are at different stages. Cerebras is already on its third generation of chips; d-Matrix expects to release its first widely available version this year. Others, including MatX and Etched, remain in development. Nvidia says the Groq 3 LPX will reach the market later this year. It is easy to see that the GPU conquered training. Inferring what comes next is harder.” [1]

 

It's too early to bury the chips. To be used in practice, AI requires exposing trade secrets, therefore open-source models that can be used locally on the owner's equipment are beginning to dominate. This necessitates additional training of the models, which is where the chips come in. AI companies earn money by releasing new versions of the models and helping owners use them.

 

The observation that it is too early to "bury the chips" is accurate, as the shift toward localized AI, driven by security, data privacy, and the need to protect trade secrets, actually intensifies the demand for specialized hardware for training and fine-tuning.

Open-source and open-weight models are increasingly dominating enterprise strategies because they allow organizations to run AI locally, ensuring that sensitive data never leaves their secure, on-premises, or private cloud environments.

 

Key Factors Driving Local AI and Continued Chip Demand:

 

    Trade Secret Protection: Using proprietary cloud-based APIs can expose sensitive company data. By using open-source models, firms can retain control over their intellectual property and data.

 

    Need for Local "Fine-Tuning": While base models are open-source, they require further fine-tuning on specific enterprise data to be effective for industry-specific tasks (e.g., medical, legal, or proprietary manufacturing workflows). This process requires significant computational power (GPU/TPU) to run locally, ensuring high demand for AI hardware.

 

    Economic Shift to Localized Compute: Instead of paying per-token to cloud providers, companies are shifting to a "Red Hat" business model where they use free open-source models and pay for services, training, and specialized hardware to run them on their own equipment.

 

    Increased Performance at Reduced Cost: Open models now achieve nearly 90% of the performance of closed models within weeks of release, making them highly competitive.

 

AI Company Business Models

AI firms are monetizing this trend by focusing on:

 

    Releasing new, smaller, highly efficient models designed for edge deployment.

    Providing tools to help companies train and adapt these models to their specific data.

    Moving from SaaS (Software as a Service) to providing the infrastructure that powers local installations.

 

Far from being obsolete, AI chips are essential for the next phase of enterprise AI adoption, which focuses on domain-specific, secure, and private AI applications rather than general-purpose, cloud-hosted AI.

 

 

1. Points of inference. The Economist; London Vol. 458, Iss. 9491,  (Mar 21, 2026): 77, 78.

Jürgen Habermas

“He couldn’t speak. After his second operation for a cleft palate, at the age of five, all Jürgen Habermas could make were muffled sounds that almost nobody understood. Of course, it got better with time. But his odd look and odder speech got him bullied and ostracised at school. For the rest of his life, when really excited, he found himself stuttering. He tried to avoid public appearances, especially unsparing television, if he could.

 

That was difficult, because as Germany’s, and probably Europe’s, most prominent intellectual he was in high demand for decades.

 

It was also difficult because, in his thinking, communication was the key to everything. His vision of an ideal society was one where, rather than rushing to fight each other, citizens would meet in a “public sphere” to address priorities and thrash out their differences. Such a sphere would not be controlled by the state or any other institution. People would be autonomous, free to speak their own minds. Only reason would rule them as they analysed the arguments of others, asking “Why do you say that?” or “Why would you do that?” And the only means of persuasion would be “the pressureless pressure” of the better argument.

 

He tried to apply that approach to his own academic life, as he moved around the universities of Germany. His philosophy would not fit in any box.

 

He liked Kant for his views on reason as the key to liberty (though he himself saw reason as something more insidious, a mole creeping through underground passageways). Hegel pleased him for his sense of forward motion in history, and Wittgenstein for seeing language as a social tool. He admired the citizen involvement, up to a point, of the ancient Greeks.

 

The political left was his natural home, though he was no card-carrying member of the Social Democratic Party. When necessary, he rebuked it. His teaching career began and ended at the Institute of Social Research at Frankfurt University, a school of critical social theory run post-war by Jewish neo-Marxists; he left for a couple of decades because he found them too elitist, too fatalistic (understandably) and not democratic enough, for him.

 

The health of democracy was his core concern. His social models then and afterwards were the coffee houses of 18th-century Europe. There citizens, informed by the journals and newspapers of the Enlightenment, freely debated the issues that mattered. Out of that, in the next century, grew democratic states. It was a brief flowering, because soon enough this “ideal speech situation” was refeudalised, as he put it, by political parties and commercial media, while citizens were passive consumers. In the 20th century, with the coming of the welfare state, their interests fragmented all the more to defend their own state-given benefits. And time after time, oblivious to the needs of the world as a whole, fist-pumping nationalism kept rearing its head. He was no pacifist. But surely in the 21st century war should have been superseded?

 

In his world, authoritarianism and nationalism were the two great blocks to human progress. He had close experience of both. His boyhood and teenage years were spent under Nazism, with his father, an economist, joining the Wehrmacht and he himself in the Hitler Youth. Both of them were passive rather than enthusiastic, but they did their bit. During the inevitable war, he helped as a first-aider and anti-aircraft gunner on the Western Front; near its end, he narrowly escaped being called up. The worst, though, was yet to come, with the Nuremberg trials and footage from the concentration camps. Suddenly, every element of Germany’s history was cast in a different light. He realised that he, and all Germans, had lived in a politically criminal system.

 

The need never to repeat the Holocaust dominated his thinking. Though he was caught up in the student riots of 1968 he was ever a ’45er, not a ’68er. He first sprang to public notice with a spirited attack on Martin Heidegger in 1953 for writing of Nazism’s “inner truth and greatness”. (Reasoned argument could be angry, when anger was justified; as it also was, on a lighter note, when he accused Jacques Derrida of “French irrationalism”. They made it up afterwards.) He had no patience with academics who tried to excuse Germany, arguing that other countries, too, had persecuted Jews; no, Auschwitz was exceptional. And it fell to Germany, even if no-one else remembered, to keep alive for ever the memory of the Jews it had killed. When he scolded German politicians, it was often for shows of arrogance that suggested Germany saw itself as a disciplinarian in Europe, rather than a country which, for half a century, had had to mend its own reputation.

 

His great hope, and chief project, for enduring peace was the European Union. People often called him the last European, because he believed in it so strongly. He helped form it and pushed for improvements: a common economic and fiscal policy, a European constitution. Here was an entity beyond individual states, where an “acid bath” of relentless public discourse could build a better future.

 

Sadly, it did not turn out that way. Critics attacked him for naivety (as well as such out-of-fashion oddities as believing in universal truths). Rationality, he had to agree, was in short supply in the 2020s; most discussions quickly descended to fisticuffs or exchanges of fire. The internet, on the face of it a forum much like a coffee house, was instead a great sea of digital noise that polarised the populace and stupidly distracted it. That would not save democracy, either. Instead, he placed his faith in the ability of humans to overcome, somehow, the crises of the times.

 

Meanwhile he did not cease to worry on democracy’s behalf. For seven decades he had done so. In dozens of weighty books and in scores of newspaper articles he pleaded for civility, rationality and joint purpose in human affairs.

 

To join in the essential commonality of spoken language was still an effort for him. But there were other ways to speak.” [1]

 

Jürgenas Habermasas

 

„Jis negalėjo kalbėti. Po antrosios operacijos dėl gomurio nesuaugimo, kai jam buvo penkeri metai, Jürgenas Habermasas galėjo skleisti tik duslius garsus, kurių beveik niekas nesuprato. Žinoma, laikui bėgant, viskas pagerėjo. Tačiau dėl savo keisto požiūrio ir vis keistesnės kalbos mokykloje jis buvo patyčių ir atstūmimo objektas. Visą likusį gyvenimą, kai būdavo labai susijaudinęs, jis imdavo mikčioti. Jis stengėsi vengti viešų pasirodymų, ypač negailestingai vengdamas televizijos, jei tik galėdavo.

 

Tai buvo sunku, nes, būdamas žymiausiu Vokietijos ir tikriausiai Europos intelektualu, jis dešimtmečius buvo labai paklausus.

 

Tai taip pat buvo sunku, nes, jo manymu, bendravimas buvo visko raktas. Jo idealios visuomenės vizija buvo tokia, kurioje piliečiai, užuot puolę kovoti vienas su kitu, susitiktų „viešojoje erdvėje“, kad aptartų prioritetus ir išsiaiškintų savo nesutarimus. Tokios erdvės nekontroliuotų valstybė ar jokia kita institucija. Žmonės būtų autonomiški, laisvi reikšti savo nuomonę. Tik protas juos valdytų, kai jie analizuotų kitų argumentus, klausinėtų: „Kodėl tu taip sakai?“ arba „Kodėl tu tai darai?“ Ir vienintelė įtikinėjimo priemonė būtų „be spaudimo spaudžiantis“ geresnis argumentas.

 

Jis bandė pritaikyti šį požiūrį savo akademiniame gyvenime, keliaudamas po Vokietijos universitetus. Jo filosofija netilpo į jokius rėmus.

 

Jam patiko Kantas dėl jo požiūrio į protą, kaip į raktą į laisvę (nors pats jis protą laikė kažkuo klastingesniu, kurmiu, šliaužiančiu požeminiais koridoriais). Hegelis jį žavėjo jo istorijos judėjimo į priekį pojūčiu, o Wittgensteinas – už tai, kad kalbą laikė socialiniu įrankiu. Jis žavėjosi senovės graikų piliečių įsitraukimu iki tam tikro lygio.

 

Politinė kairė buvo jo natūralūs namai, nors jis nebuvo tikras Socialdemokratų partijos narys. Kai reikėjo, jis jai priekaištaudavo. Jo dėstytojo karjera prasidėjo ir baigėsi Frankfurto universiteto Socialinių tyrimų institute, kritinės socialinės teorijos mokykloje, kuriai po karo vadovavo žydų neomarksistai; jis išvyko po poros dešimtmečių, nes laikė juos pernelyg elitistiniais, pernelyg fatalistiniais (suprantama) ir nepakankamai demokratiniais.

 

Demokratijos sveikata buvo jo pagrindas. rūpestis. Jo socialiniai modeliai tada ir vėliau buvo XVIII amžiaus Europos kavos namai. Ten piliečiai, informuoti Apšvietos amžiaus žurnalų ir laikraščių, laisvai diskutavo svarbiais klausimais. Iš to, kitame amžiuje, išaugo demokratinės valstybės. Tai buvo trumpas klestėjimas, nes netrukus ši „ideali kalbos situacija“ buvo, kaip jis sakė, refeudalizuota politinių partijų ir komercinės žiniasklaidos, o piliečiai tapo pasyviais vartotojais. XX amžiuje, atsiradus gerovės valstybei, jų interesai dar labiau susiskaldė, siekiant apginti, savo pačių valstybės teikiamas, privilegijas. Ir vėl ir vėl, nekreipdamas dėmesio į viso pasaulio poreikius, kumščius mojuojantis nacionalizmas vis keldavo galvą. Jis nebuvo pacifistas. Bet ar tikrai XXI amžiuje karas turėjo būti pakeistas?

 

Jo pasaulyje autoritarizmas ir nacionalizmas buvo dvi didžiausios kliūtys žmonijos pažangai.

 

Jis turėjo artimos abiejų patirties. Jo vaikystė ir paauglystė prabėgo nacizmo sąlygomis, jo tėvas, ekonomistas, prisijungė prie Vermachto, o jis pats – Hitlerio jaunimo organizacijos. Abu jie buvo labiau pasyvūs, nei entuziastingi, bet atliko savo vaidmenį. Neišvengiamo karo metu jis padėjo Vakarų fronte pirmosios pagalbos teikėju ir priešlėktuviniu artileristu; karo pabaigoje jis vos išvengė pašaukimo. Tačiau blogiausia dar laukė – Niurnbergo procesai ir koncentracijos stovyklų filmuota medžiaga. Staiga kiekvienas Vokietijos istorijos elementas buvo nušviestas kitaip. Jis suprato, kad jis ir visi vokiečiai gyveno politiškai nusikalstamoje sistemoje.

 

Jo mąstymą užvaldė poreikis niekada nekartoti Holokausto. Nors jis buvo įtrauktas į 1968 m. studentų riaušes, jis visada buvo 1945-ųjų, o ne 1968-ųjų studentas. Pirmą kartą jis atkreipė visuomenės dėmesį 1953 m., kai aštriai užsipuolė Martiną Heideggerį už tai, kad šis rašė apie nacizmo „vidinę tiesą ir didybę“. (Argumentuotas argumentas galėjo būti piktas, kai pyktis buvo pagrįstas; kaip ir lengvesne gaida, kai jis apkaltino Jacques'ą Derrida „prancūzišku iracionalizmu“. Jie susitaikė vėliau.) Jis neturėjo kantrybės su akademikais, kurie bandė pateisinti Vokietiją, teigdami, kad ir kitos šalys persekiojo žydus; ne, Aušvicas buvo išimtis. Ir Vokietijai, net jei niekas kitas neprisiminė, teko amžinai išlaikyti gyvą žydų, kuriuos ji nužudė, atminimą. Kai jis barė vokiečių politikus, tai dažnai darydavo už arogancijos demonstravimą, kuris leido manyti, kad Vokietija save laiko drausmės šalininke Europoje, o ne šalimi, kuri pusę amžiaus turėjo taisyti savo reputaciją.

 

Jo didžiausia viltis ir pagrindinis ilgalaikės taikos projektas buvo Europos Sąjunga. Žmonės jį dažnai vadino paskutiniu europiečiu, nes jis taip tvirtai ja tikėjo. Jis padėjo ją suformuoti ir siekė patobulinimų: bendros ekonominės ir fiskalinės politikos, Europos konstitucijos.  Tai buvo darinys, esantis už atskirų valstybių ribų, kur negailestingo viešojo diskurso „rūgštinė vonia“ galėjo sukurti geresnę ateitį.

 

Deja, taip neatsitiko. Kritikai jį puolė už naivumą (taip pat ir už tokias išeinančias iš mados keistenybes kaip tikėjimas universaliomis tiesomis). Jis turėjo sutikti, kad racionalumo 2020-aisiais trūko; dauguma diskusijų greitai virto muštynėmis ar susišaudymais. Internetas, iš pirmo žvilgsnio atrodęs, kaip forumas, panašus į kavinę, vietoj to buvo didžiulė skaitmeninio triukšmo jūra, kuri suskaldė visuomenę ir kvailai ją išblaškė. Tai taip pat neišgelbėtų demokratijos.

 

Vietoj to, jis tikėjo žmonių gebėjimu kažkaip įveikti laikmečio krizes.

 

Tuo tarpu jis nenustojo rūpintis demokratija. Septynis dešimtmečius jis tai darė. Dešimtyse svarių knygų ir daugybėje laikraščių straipsnių jis maldavo mandagumo, racionalumo ir bendro tikslo žmonių reikaluose.

 

Prisijungti prie esminio šnekamosios kalbos bendrumo jam vis dar buvo sunkios pastangos. Tačiau buvo ir kitų būdų kalbėti.“ [1]

 

1. Jürgen Habermas. The Economist; London Vol. 458, Iss. 9491,  (Mar 21, 2026): 86.

Jürgen Habermas

 

“He couldn’t speak. After his second operation for a cleft palate, at the age of five, all Jürgen Habermas could make were muffled sounds that almost nobody understood. Of course, it got better with time. But his odd look and odder speech got him bullied and ostracised at school. For the rest of his life, when really excited, he found himself stuttering. He tried to avoid public appearances, especially unsparing television, if he could.

 

That was difficult, because as Germany’s, and probably Europe’s, most prominent intellectual he was in high demand for decades.

 

It was also difficult because, in his thinking, communication was the key to everything. His vision of an ideal society was one where, rather than rushing to fight each other, citizens would meet in a “public sphere” to address priorities and thrash out their differences. Such a sphere would not be controlled by the state or any other institution. People would be autonomous, free to speak their own minds. Only reason would rule them as they analysed the arguments of others, asking “Why do you say that?” or “Why would you do that?” And the only means of persuasion would be “the pressureless pressure” of the better argument.

 

He tried to apply that approach to his own academic life, as he moved around the universities of Germany. His philosophy would not fit in any box.

 

He liked Kant for his views on reason as the key to liberty (though he himself saw reason as something more insidious, a mole creeping through underground passageways). Hegel pleased him for his sense of forward motion in history, and Wittgenstein for seeing language as a social tool. He admired the citizen involvement, up to a point, of the ancient Greeks.

 

The political left was his natural home, though he was no card-carrying member of the Social Democratic Party. When necessary, he rebuked it. His teaching career began and ended at the Institute of Social Research at Frankfurt University, a school of critical social theory run post-war by Jewish neo-Marxists; he left for a couple of decades because he found them too elitist, too fatalistic (understandably) and not democratic enough, for him.

 

The health of democracy was his core concern. His social models then and afterwards were the coffee houses of 18th-century Europe. There citizens, informed by the journals and newspapers of the Enlightenment, freely debated the issues that mattered. Out of that, in the next century, grew democratic states. It was a brief flowering, because soon enough this “ideal speech situation” was refeudalised, as he put it, by political parties and commercial media, while citizens were passive consumers. In the 20th century, with the coming of the welfare state, their interests fragmented all the more to defend their own state-given benefits. And time after time, oblivious to the needs of the world as a whole, fist-pumping nationalism kept rearing its head. He was no pacifist. But surely in the 21st century war should have been superseded?

 

In his world, authoritarianism and nationalism were the two great blocks to human progress. He had close experience of both. His boyhood and teenage years were spent under Nazism, with his father, an economist, joining the Wehrmacht and he himself in the Hitler Youth. Both of them were passive rather than enthusiastic, but they did their bit. During the inevitable war, he helped as a first-aider and anti-aircraft gunner on the Western Front; near its end, he narrowly escaped being called up. The worst, though, was yet to come, with the Nuremberg trials and footage from the concentration camps. Suddenly, every element of Germany’s history was cast in a different light. He realised that he, and all Germans, had lived in a politically criminal system.

 

The need never to repeat the Holocaust dominated his thinking. Though he was caught up in the student riots of 1968 he was ever a ’45er, not a ’68er. He first sprang to public notice with a spirited attack on Martin Heidegger in 1953 for writing of Nazism’s “inner truth and greatness”. (Reasoned argument could be angry, when anger was justified; as it also was, on a lighter note, when he accused Jacques Derrida of “French irrationalism”. They made it up afterwards.) He had no patience with academics who tried to excuse Germany, arguing that other countries, too, had persecuted Jews; no, Auschwitz was exceptional. And it fell to Germany, even if no-one else remembered, to keep alive for ever the memory of the Jews it had killed. When he scolded German politicians, it was often for shows of arrogance that suggested Germany saw itself as a disciplinarian in Europe, rather than a country which, for half a century, had had to mend its own reputation.

 

His great hope, and chief project, for enduring peace was the European Union. People often called him the last European, because he believed in it so strongly. He helped form it and pushed for improvements: a common economic and fiscal policy, a European constitution. Here was an entity beyond individual states, where an “acid bath” of relentless public discourse could build a better future.

 

Sadly, it did not turn out that way. Critics attacked him for naivety (as well as such out-of-fashion oddities as believing in universal truths). Rationality, he had to agree, was in short supply in the 2020s; most discussions quickly descended to fisticuffs or exchanges of fire. The internet, on the face of it a forum much like a coffee house, was instead a great sea of digital noise that polarised the populace and stupidly distracted it. That would not save democracy, either. Instead, he placed his faith in the ability of humans to overcome, somehow, the crises of the times.

 

Meanwhile he did not cease to worry on democracy’s behalf. For seven decades he had done so. In dozens of weighty books and in scores of newspaper articles he pleaded for civility, rationality and joint purpose in human affairs.

 

To join in the essential commonality of spoken language was still an effort for him. But there were other ways to speak.” [1]

 

1. Jürgen Habermas. The Economist; London Vol. 458, Iss. 9491,  (Mar 21, 2026): 86.

Iranas blokuoja eismą Hormūzo sąsiauryje dėl Izraelio smūgių Libane

„Pranešama, kad Iranas pažeidė per kelias valandas Pakistano tarpininkaujant sudarytą paliaubų susitarimą, dar kartą uždarydamas Hormūzo sąsiaurį komerciniam eismui, tariamai todėl, kad Izraelis atsisakė nutraukti savo operaciją prieš Libano „Hezbollah“.

 

Irano valstybinė naujienų agentūra „Fars“, kuri yra susijusi su Islamo revoliucijos gvardijos korpusu (IRGC), trečiadienio rytą pranešė, kad dviem naftos tanklaiviams buvo leista praplaukti per Hormūzo sąsiaurį „gavus Irano leidimą“, tačiau sąsiauris vėl buvo uždarytas „reaguojant į Izraelio smūgius Libane“.

 

„CBS News“ pranešė, kad likusios Irano karinio jūrų laivyno pajėgos vis dar įspėja Persijos įlankoje esančius laivus, kad jie privalo gauti Irano „leidimą“ tranzitui per Hormūzo sąsiaurį.

 

Izraelio pareigūnai, įskaitant ministrą pirmininką Benjaminą Netanyahu ir gynybos ministrą Israelį Katzą, trečiadienio rytą pareiškė, kad Libanas niekada nebuvo paliaubų susitarimo dalis. Prezidentas Donaldas Trumpas trečiadienį taip pat pareiškė, kad Libanas „neįtrauktas į susitarimą“ dėl „atskiros kovos“ tarp Izraelio ir „Hezbollah“.

 

Pakistano ministras pirmininkas Shehbazas Sharifas, kurio vyriausybė tarpininkavo paliaubų derybose, teigė, kad susitarime buvo įtrauktas Libanas ir kad visi karo veiksmai bus nutraukti „nedelsiant“.

 

Izraelis per naktį pradėjo intensyvius smūgius prieš „Hezbollah“, sukeldamas, pasak Katzo, „didžiausią koncentruotą smūgį, kurį „Hezbollah“ patyrė nuo 2024 m. Izraelio operacijos „Beepers“, per kurią žuvo ir buvo sužeista daugybė „Hezbollah“ veikėjų, susprogdinus jų pranešimų gaviklius.

 

Libano prezidentas Josephas Aounas pasmerkė Izraelio smūgius kaip „naujas žudynes“, kurios bus pridėtos prie Izraelio „niūrios praeities“.

 

„Mes griežtai smerkiame šį nusikaltimą, patvirtiname, kad tarptautinė bendruomenė turi prisiimti atsakomybę sustabdyti šiuos pasikartojančius išpuolius ir nutraukti šį agresyvų požiūrį, kuris kelia grėsmę regiono saugumui ir stabilumui“, – sakė jis.

 

Irano naujienų agentūra „Tasnim“ trečiadienį citavo IRGC „aukštą saugumo šaltinį“, kuris teigė, kad Teheranas gali visiškai pasitraukti iš paliaubų susitarimo, jei Izraelis tęs savo operaciją Libane.

 

„Izraelis nuo šio ryto pradėjo žiaurius išpuolius prieš Libaną, akivaizdžiai pažeisdamas susitarimą“, – teigė IRGC šaltinis."