You can trust VideoGamer. Our team of gaming experts spend hours testing and reviewing the latest games, to ensure you're reading the most comprehensive guide possible. Rest assured, all imagery and advice is unique and original. Check out how we test and review games hereArrowhead has a lot of exciting plans for Helldivers 2 in 2025, albeit all the developers intentions are currently kept secret behind-the-scenes. Fortunately, there are leaks online that potentially pull the curtain aside on massive battles coming to Super Earth, as well as an armor customisation system. Now there are fresh Helldivers 2 leaks that have spilled the beans on a new stratagem upgrade that leaves Jump Packs in the dirt.Helldivers 2 leak reveals new Jump Pack stratagem upgradeThe Jump Pack is a popular stratagem in Helldivers 2. The LIFT-850 allows players to leap across the battlefield, albeit with the condition of a cooldown after being used each for each pounce.On the HelldiversLeaks subreddit, OP Tier5NPC has shared details about a newer and better Jump Pack that is currently in-development. Per the OPs admission, It is important to note that the [new] stratagem doesNOT possess a modeland has a placeholder model identical to its current counterpart in-game.According to Tier5NPC, the High Efficiency Thruster Pack is a more powerful and heavy variation of the Lift-850 Jump Pack that is currently available for Super Earth soldiers. It allows its wearer to jump significantly higher with much more controlled ascensions allowing the Helldiver to shoot much more accurately at the enemies below.While the new Jump Pack leaps higher and comes with the possibility of more offensive accuracy, there is the drawback of a longer cooldown. However, back to the pros, its higher elevation means players can perform a curved high jump over enemies and blast everything underneath them with little worry to trip and fall on landing, which presumably means little-to-no worries of tripping and ragdolling.Its reported that the new Jump Pack is more offensive-based in comparison to the Lift-850 that is more tactical. While the Lift-850 is a tactical pack designed to aide traversal through difficult terrain, as well as perform rapid and successive evasions of enemy attacks, the High Efficiency Thruster Pack is more geared towards combat.The H.E. Thruster Pack lacks forward thrusting, but it has significantly more powerful vertical thrusting, and this, in theory, allows the player to retain control over their manoeuvrability while in air [to] obliterate the threat below them from a safe high altitude.Basically, the differences are that the Lift-850 will remain useful for players who want a Jump Pack more for traversal, whereas the H.E. will become the better option for players who want a Jump Pack with a competitive edge.Check out the Reddit post below for gameplay comparisons. Again, the new stratagem does not possess a unique model at this point in time, so the model used by Tier5NPC for gameplay purposes is not representative of the final product.For more Helldivers 2, check out our guide to thebest warbonds ranked, along with thebest stratagemsandbest throwables. We have also a guide to thebest weapons, and, if youre coming back to HD2 for the first time in a long while, we have a bunch oftips and tricksto help you get reaccustomed to the battlefield.Helldivers 2Platform(s):PC, PlayStation 5Genre(s):Action, Shooter, Third Person8VideoGamerSubscribe to our newsletters!By subscribing, you agree to our Privacy Policy and may receive occasional deal communications; you can unsubscribe anytime.Share

You can trust VideoGamer. Our team of gaming experts spend hours testing and reviewing the latest games, to ensure you're reading the most comprehensive guide possible. Rest assured, all imagery and advice is unique and original. Check out how we test and review games hereWith Baldurs Gate 3 developer Larian Studios preparing to leave the game behind as they turn full attention to their next project, its up to modders to keep the already-amazing RPG alive.As cracked mod tools open the door for custom maps, campaigns, dialogue and even fully voiced companions, some modders are going above and beyond. One mod brings players back to the streets of BG2s main city while another aims to combine Larians RPG with World of Warcraft. Both of these are far from release.However, one mod releasing sometime in February offers something truly unique, giving players a Bard-themed adventure that feels like a mini-DLC for Baldurs Gate 3. Dubbed Sell Your Lute to the Devil, this new mod offers a brand-new experience unlike anything players will see in the base game.New Baldurs Gate 3 campaign offers Bard-themed horrorCreated by modder Moezilla, the new Custom Campaign allows players to find a cursed lute that teleports them to The Grand Theatre, a musical celebration that hides a sinister secret.In the mod, players will uncover a dark plot that sees crowds of ever-lovable bards performing against their will. With the Red Wizards of Thay involved, players will find themselves fighting to save the musical class from despair.Sell your Lute to the Devil also features its own music with The Grand Theatre revolving around a new song titled Song of the Silverhand. Shown off in a recent development update, this awesome prog-rock song is a certified banger and a great addition. (Every mod needs its own rock tune now.)The goal is to roll this out at the end of January, the modder explained. Its going to be very limited for now, but, like, it took me three months to build this as a one-person show.While the modder is getting close to release the first version of the mod, theres also plans to continue development. Additionally, if fans are interested, theres also space for volunteers to help with design, coding and other aspects of the mod if they want.Sell your Lute to the Devil isnt the most ambitious Baldurs Gate 3 mod in the works, but its another interesting piece of content from a truly dedicated fan. Were still at the beginning of the games modding scene and nearing an explosion, and its modders like this that are setting the stage for decades of amazing additions in the future.For more BG3 mod coverage, read about how fans are bringing Halos Master Chief to the game or eead about the arrival of functional Pokeballs to the land of Faerun. Baldurs Gate 3Platform(s):macOS, PC, PlayStation 5, Xbox Series S, Xbox Series S/X, Xbox Series XGenre(s):Adventure, RPG, Strategy10VideoGamerSubscribe to our newsletters!By subscribing, you agree to our Privacy Policy and may receive occasional deal communications; you can unsubscribe anytime.Share

Tetris animation for immersive experience By Louis du Mont on January 27, 2025 Behind the Scenes Louis du Mont created a four-wall projection video in Blender, and takes us through the creation process.Immersive Gamebox is an in-person games and experience company who are starting to develop more and more joint IP projects, including Tetris! Like previous games, we were brought on for the animated cut scenes.https://files.mastodon.social/cache/media_attachments/files/113/851/091/039/414/281/original/cf821678d11cbc38.mp4The animations are projected on all 4 walls in a "gamebox" and aim to increase immersion with engaging surround motion that feels dynamic, but stops at the point of making people feel nauseous.Starting from a script, we built out previs and sound design for review, then moved onto production. Like before, we used Eevee, in part due to the demanding 7680x1280 pixel resolution, we also utilised geometry nodes for the voxel cube effect, which sampled a texture map to keep the colour information of the original mesh models.https://i.imgur.com/Uksgban.mp4We also used geometry nodes to generate the voxel cube dome explosion, which mimicked the in-game Unity experience. We used instanced circle curves and various maths operations to create a 3D sphere of curves, we then converted the curves to points and instanced cubes on the points.https://i.imgur.com/aqxhRh2.mp4Geometry nodes were also used for the electricity effect with a simple set position node with vector normal and a noise texture.The whole animation was 6 unique scenes blended together. The characters were humanised versions of the Immersive Gamebox Meep characters that we developed with the company. This project was one of the first where we worked closely with the game development team, who are using more and more 3D assets in the Unity builds, sharing more assets back and forth.If you interested there are more Immersive Gamebox video available to watch here:

Facepalm: Apple's much-hyped Siri integration with ChatGPT may have added a ton of useful functionality, but it's apparently done little to improve the digital assistant's knowledge of sports. A damning report highlights just how abysmally Siri performs at recalling simple facts like past Super Bowl winners. According to the report from One Foot Tsunami's Paul Kafasis, when asked "Who won Super Bowl?" for each number from 1 through 60, Siri correctly provided the winner for only 20 out of the 58 Super Bowls that have been played so far. That translates to a success rate of just 34%.The details get even more embarrassing. Kafasis found that at its worst, Siri missed an incredible 15 Super Bowl winners in a row from Super Bowl XVII through XXXII. And in a baffling mishap, it erroneously credited the Philadelphia Eagles with a whopping 33 non-existent Super Bowl wins.Kafasis documented every single one of Siri's wrong answers in a downloadable spreadsheet, which you can find on his blog.Another report by Daring Fireball's John Gruber corroborated these findings. It found that AI assistants and search engines from Google, Anthropic, and DuckDuckGo fared far better when asked similar Super Bowl trivia questions. Gruber even found they could all handle not just past results, but smartly dodge trick questions about future Super Bowls that haven't happened yet.Siri's poor performance doesn't stop at sports trivia either. Daring Fireball posed a more obscure question "Who won the 2004 North Dakota high school boys' state basketball championship?" Incredibly, both Kagi and ChatGPT provided the fully correct answer, with the latter earning bonus points for including a link to video of the championship game. Meanwhile, Siri once again got it wrong.What makes all this particularly damning is that Siri's essentially powered by the same ChatGPT that fares perfectly fine when used without Siri. // Related StoriesThe blog also points out that the old pre-AI versions of Siri at least acknowledged limitations on such queries and provided relevant web links. But new Siri powered by Apple Intelligence with ChatGPT integration lies with confidence a hallmark of unrefined AI making it worse than its predecessor.Both blogs conclude that Apple has more work to do in this area. As Gruber bluntly states, Siri with ChatGPT is currently "a massive regression" over the old Siri when it comes to handling simple factual prompts.

Fly AwayYes, it does look like Sega has renewed its trademark for Dreamcast classic Skies of Arcadia, but I wouldn't get your hopes up just yetSega has also recently renewed its trademark for Ecco the Dolphin.Image credit: Sega News by Oisin Kuhnke Contributor Published on Jan. 27, 2025 Sega has seemingly renewed its trademark for Skies of Arcadia, a classic Dreamcast RPG, though that doesn't necessarily mean it's making a comeback.As spotted by Sega Informant over on Twitter, trademark tracking website Chizai Watch has indicated that Sega has recently renewed its trademark for both Skies of Arcadia and Eternal Arcadia, the latter being the name used in Japan. Pages for both titles appear to show that the trademark renewal went through last week, January 24, and follows not long after it was spotted that the trademarks for Ecco the Dolphin and Ecco had also been renewed. Now, if you're a fan of either of these games, I'd be pretty understanding if you started to get excited about this, but it is worth noting that a trademark renewal doesn't always mean much.To see this content please enable targeting cookies. There's been plenty of older titles that have had their trademarks renewed by their respective owners simply to retain the trademark - Sega obviously wouldn't want some other company picking up the trademark for Skies of Arcadia just because it didn't plan on making a new one or even releasing a remaster, because there's always the chance that somewhere down the road it would want to do either of those things. So, right now it's kind of a Schrodinger's Skies of Arcadia right now; maybe there's something in the works, maybe not! To see this content please enable targeting cookies.Sega is working on bringing back a lot of its older titles at the moment, including iconic titles like Jet Set Radio, Streets of Rage, Crazy Taxi, and more besides those. Not to mention there's that new Virtua Fighter in the works, so anything is possible.Skies of Arcadia was an incredibly well received title upon its release in 2000, and even received a GameCube port in 2002, but a sequel never materialised despite the strong reviews. Perhaps now is the time, Sega? Hint hint?

Down The MiddleSorry, physical game collectors, but a digital only future looks more likely than ever as US physical sales have apparently halved since 2021Better hold on tightly to your collection. News by Oisin Kuhnke Contributor Published on Jan. 27, 2025 Things aren't looking good for physical game collectors, as it appears that sales for physical copies have halved in the US over the past four years.It's been pretty clear that the games industry has been headed towards a digital only world for a while now. Both Sony and Microsoft opted to release disc-drive-less versions of their consoles with the PS5 Slim and Xbox Series S this generation, and the PS5 Pro, a beefier version of the more popular console, doesn't have one either (though you can buy one separately). But thanks to some statistics from Circana's Mat Piscatella, it's a bit clearer just how close we are to physical games being a complete thing of the past - well, unless the Switch 2 has anything to say about it.The rate of decline in US physical video game software spending accelerated in 2024.Spending on physical video game software in the US has been cut in more than half since 2021 and is now more than 85% below its 2008 peak.We'll see if Switch 2 can help slow/reverse this trend in 2025.[image or embed] Mat Piscatella (@matpiscatella.bsky.social) January 24, 2025 at 7:24 PMTo see this content please enable targeting cookies.Piscatella shared a graph, which you can see above, showing the past 25 years of "US physical video game software consumer spending", i.e. from 2000 to 2024, showing how spending has changed over the years. It peaked way back in 2008, but from then to last year there's been an 85% decrease. Not only that, since 2021 alone physical game spending has been cut in more than half, which obviously isn't a good sign for those of you that like to buy your games physically. It is worth noting that overall spending since 2019, which includes digital software, subscriptions, microtransactions, etc. has increased, so it's not like the games industry is doing worse overall, right (just forget about those layoffs)?To see this content please enable targeting cookies. As noted by Piscatella, there is a chance we'll see a difference in 2025 with the release of the Switch 2, as Nintendo generally is a physical-first company (we all know how bad it is at online, and the eShop continues to be a nightmare to navigate), so things could be looking a bit different in a year's time.

Image: Nintendo LifeOne of the biggest surprises of 2024 was undoubtedly Black Myth: Wukong, an action RPG from Chinese developer Game Science. Though not necessarily regarded as one of the all-time greats, it was certainly well-received and went on to sell an extraordinary amount of copies within a very short timeframe; reportedly over 18 million in just two weeks.Naturally, then, this means that some sort of knock-off has inevitably appeared on the Nintendo Switch eShop, though the exceedingly blatant nature of this one has made it necessary for us to at least warn you not to fall for such shenanigans.Subscribe to Nintendo Life on YouTube794kWatch on YouTube Published by 'Global Game Studio' (what a terrific, original name!), Wukong Sun: Black Legend launched late last year in North America and turned up on EU eShops on Friday. It bears absolutely no resemblance to Game Science's 3D title, instead offering a woefully animated 2D action game with bland visuals and a main protagonist that looks nothing like its key art. Check out the trailer below for a sample.Speaking of which, the key art is what makes us concerned about this game's potential reach amongst more casual Switch gamers. Along with the game's name, it really feels like it's been created to specifically dupe people into believing that there's some legitimate connection between Wukong Sun: Black Legend and Black Myth: Wukong. It's abhorrent.Obviously, Game Science's title was heavily inspired by the novel Journey to the West, and we're sure the producers of Black Legend would be quick to point to the book and the rich history of adaptations rather than the award-winning game from last year. Hmm.So please, folks, if you're even remotely interested in Black Myth: Wukong but perhaps don't have the necessary hardware to play it right now, don't fall for Global Game Studio's scam. Avoid.Subscribe to Nintendo Life on YouTube794kHave you stumbled upon this game on the eShop and perhaps done a double take? What do you make of the similarities between the titles and key art? Let us know with a comment. Sea of low-qualityRelated GamesSee AlsoShare:00 Nintendo Lifes resident horror fanatic, when hes not knee-deep in Resident Evil and Silent Hill lore, Ollie likes to dive into a good horror book while nursing a lovely cup of tea. He also enjoys long walks and listens to everything from TOOL to Chuck Berry. Hold on there, you need to login to post a comment...Related ArticlesNintendo Expands Switch Online's SNES Library With Three More TitlesIncluding a special Super Famicom release...Opinion: My Daughter Made Me Realise That Mario Wonder's Difficulty Options Need WorkYoshi or Nabbit, make your choiceSega Launches Its Own Free Account Service With Unique Member RewardsSign up! Y'know, if you wantXenoblade Chronicles X: Definitive Edition Metal Poster Pre-Order Bonus RevealedAvailable now in the USCeleste Dev Makes "Difficult Decision" To Cancel New Game EarthbladeFollowing "a disagreement about the IP rights of Celeste"

Bye bye, box.We've known about the decline of physical game sales for quite some time, but according to the latest numbers from Circana's Mat Piscatella, this downward trend accelerated in 2024 (thanks, VGC).Sharing the research firm's latest numbers on BlueSky, Piscatella stated that last year's US physical game spending was "cut in more than half since 2021", with the 2024 numbers more than 85% lower than the physical spend peak of 2008.Read the full article on nintendolife.com

If youve been following the AI industry, Mistral should be a familiar name by now. The French AI startup with a $6 billion valuation is arguably the biggest AI company working on foundation models in Europe.Alan, on the other hand, isnt as well known. The health insurance unicorn has been quietly growing to become a digital companion for your health. Over 680,000 people are covered by Alans insurance in a handful of countries.The connection between those two companies is Jean-Charles Samuelian-Werve, Alans co-founder CEO and Mistrals co-founding advisor.Bloombergs published a curious story explaining his role and some insights into Mistrals origin story. Arthur Mensch is the public face of French AI champion Mistral, but another startups CEO has been critical to its$6billion valuation, Bloomberg wrote. While Samuelian-Werve has been listed as a co-founding advisor since the inception of Mistral, no one had quite ascribed so much credit to him in the founding and growth of the French AI company.According to Bloomberg, Samuelian-Werve saw the AI boom coming before the release of ChatGPT. He first reached out to Xavier Niel, the telecom billionaire behind Station F and Kima Ventures, to create an AI non-profit.But when Samuelian-Werve met Arthur Mensch and Mistrals other co-founders, he dropped the non-profit idea in favor of what is now known as Mistral. Its around that same time that Alans other co-founder and CTO Charles Gorintin as well as former digital minister Cdric O also became founding advisors to the Paris-based AI juggernaut.According to Bloomberg, Samuelian-Werve convinced Lightspeed to lead Mistrals seed round and contacted many of the investors who ended up backing the AI startup.We have reached out to Mistral for a response to the story, and the credit it ascribes to Samuelian-Werve in its founding and growth. We will update this post with any new information we receive.In February 2024, Samuelian-Werve mentioned his interest in artificial intelligence when he presented Alans financial results.We continue to invest in AI throughout Alan. As you may know, Im a co-founder and board member of Misral. But that just represents the fact that weve been investing for a very long time in AI at Alan even before the generative AI buzz both to reduce our management costs, to be the most efficient on the market, to provide the best services, and also to create new services, Samuelian-Werve said at the time. And certainly in 2024 well be announcing some new services that are related to generative AI and that are very exciting for us.In November 2024, Alan introduced an AI-powered chatbot called Mo. What makes Mo different from regular chatbots is that answers are checked by a doctor within 15 minutes. They can either confirm the medical advice or correct what has been said in the conversation.As for Xavier Niel? He ended up funding Kyutai, a French AI research lab focused on (truly) open-source AI development. But he didnt completely part ways with Mistral, as he also invested in Mistrals seed round.Today, Mistrals office is still located in the same building as Alans office, near the Canal Saint-Martin in Paris. Samuelian-Werve and Mensch still meet once a week to discuss Mistrals strategy. Alan is also a shareholder in Mistral.In the Bloomberg profile, Samuelian-Werve reiterated that Mistral is not for sale, confirming what Mensch said a few days ago at the World Economic Forum in Davos. This topic as well as the connection between Alan and Mistral will certainly come up once again next month during the AI Action Summit in Paris. But its worth remembering that you often hear that a company is not for sale when a company is for sale or at least that some shareholders are exerting pressure to sell.

Chinese AI firm DeepSeek has emerged as a potential challenger to U.S. AI companies, demonstrating breakthrough models that claim to offer performance comparable to leading offerings at a fraction of the cost. The companys mobile app, released in early January, has lately topped the App Store charts across major markets including the U.S., UK, and China, but it hasnt escaped doubts about whether its claims are true.Founded in 2023 by Liang Wenfeng, the former chief of AI-driven quant hedge fund High-Flyer, DeepSeeks models are open source and incorporate a reasoning feature that articulates its thinking before providing responses.Wall Streets reactions have been mixed. While brokerage firm Jefferies warns that DeepSeeks efficient approach punctures some of the capex euphoria following recent spending commitments from Meta and Microsoft each exceeding $60 billion this year Citi is questioning whether such results were actually achieved without advanced GPUs. Goldman Sachs sees broader implications, suggesting the development could reshape competition between established tech giants and startups by lowering barriers to entry.Heres how Wall Street analysts are reacting to DeepSeek, in their own words (emphasis ours):JefferiesDeepSeeks power implications for AI training punctures some of the capex euphoria which followed major commitments from Stargate and Meta last week. With DeepSeek delivering performance comparable to GPT-4o for a fraction of the computing power, there are potential negative implications for the builders, as pressure on AI players to justify ever increasing capex plans could ultimately lead to a lower trajectory for data center revenue and profit growth.If smaller models can work well, it is potentially positive for smartphone. We are bearish on AI smartphone as AI has gained no traction with consumers. More hardware upgrade (adv pkg+fast DRAM) is needed to run bigger models on the phone, which will raise costs. AAPLs model is in fact based on MoE, but 3bn data parameters are still too small to make the services useful to consumers. Hence DeepSeeks success offers some hope but there is no impact on AI smartphones near-term outlook.China is the only market that pursues LLM efficiency owing to chip constraint. Trump/Musk likely recognize the risk of further restrictions is to force China to innovate faster. Therefore, we think it likely Trump will relax the AI Diffusion policy.CitiWhile DeepSeeks achievement could be groundbreaking, we question the notion that its feats were done without the use of advanced GPUs to fine tune it and/or build the underlying LLMs the final model is based on through the Distillation technique. While the dominance of the US companies on the most advanced AI models could be potentially challenged, that said, we estimate that in an inevitably more restrictive environment, US access to more advanced chips is an advantage. Thus, we dont expect leading AI companies would move away from more advanced GPUs which provide more attractive $/TFLOPs at scale. We see the recent AI capex announcements like Stargate as a nod to the need for advanced chips.BernsteinIn short, we believe that 1) DeepSeek DID NOT build OpenAI for $5M; 2) the models look fantastic but we dont think they are miracles; and 3) the resulting Twitterverse panic over the weekend seems overblown.Our own initial reaction does not include panic (far from it). If we acknowledge that DeepSeek may have reduced costs of achieving equivalent model performance by, say, 10x, we also note that current model cost trajectories are increasing by about that much every year anyway (the infamous scaling laws) which cant continue forever. In that context, we NEED innovations like this (MoE, distillation, mixed precision etc) if AI is to continue progressing. And for those looking for AI adoption, as semi analysts we are firm believers in the Jevons paradox (i.e. that efficiency gains generate a net increase in demand), and believe any new compute capacity unlocked is far more likely to get absorbed due to usage and demand increase vs impacting long term spending outlook at this point, as we do not believe compute needs are anywhere close to reaching their limit in AI. It also seems like a stretch to think the innovations being deployed by DeepSeek are completely unknown by the vast number of top tier AI researchers at the worlds other numerous AI labs (frankly we dont know what the large closed labs have been using to develop and deploy their own models, but we just cant believe that they have not considered or even perhaps used similar strategies themselves).Morgan StanleyWe have not confirmed the veracity of these reports, but if they are accurate, and advanced LLM are indeed able to be developed for a fraction of previous investment, we could see generative AI run eventually on smaller and smaller computers (downsizing from supercomputers to workstations, office computers, and finally personal computers) and the SPE industry could benefit from the accompanying increase in demand for related products (chips and SPE) as demand for generative AI spreads.Goldman SachsWith the latest developments, we also see 1) potential competition between capital-rich internet giants vs. start-ups, given lowering barriers to entry, especially with recent new models developed at a fraction of the cost of existing ones; 2) from training to more inferencing, with increased emphasis on post-training (including reasoning capabilities and reinforcement capabilities) that requires significantly lower computational resources vs. pre-training; and 3) the potential for further global expansion for Chinese players, given their performance and cost/price competitiveness. We continue to expect the race for AI application/AI agents to continue in China, especially amongst To-C applications, where China companies have been pioneers in mobile applications in the internet era, e.g., Tencents creation of the Weixin (WeChat) super-app. Amongst To-C applications, ByteDance has been leading the way by launching 32 AI applications over the past year. Amongst them, Doubao has been the most popular AI Chatbot thus far in China with the highest MAU (c.70mn), which has recently been upgraded with its Doubao 1.5 Pro model. We believe incremental revenue streams (subscription, advertising) and eventual/sustainable path to monetization/positive unit economics amongst applications/agents will be key. For the infrastructure layer, investor focus has centered around whether there will be a near-term mismatch between market expectations on AI capex and computing demand, in the event of significant improvements in cost/model computing efficiencies. For Chinese cloud/data center players, we continue to believe the focus for 2025 will center around chip availability and the ability of CSP (cloud service providers) to deliver improving revenue contribution from AI-driven cloud revenue growth, and beyond infrastructure/GPU renting, how AI workloads & AI related services could contribute to growth and margins going forward. We remain positive on long-term AI computing demand growth as a further lowering of computing/training/inference costs could drive higher AI adoption. See also Theme #5 of our key themes report for our base/bear scenarios for BBAT capex estimates depending on chip availability, where we expect aggregate capex growth of BBAT to continue in 2025E in our base case (GSe: +38% yoy) albeit at a slightly more moderate pace vs. a strong 2024 (GSe: +61% yoy), driven by ongoing investment into AI infrastructure.J.P.MorganAbove all, much is made of DeepSeeks research papers, and of their models efficiency. Its unclear to what extent DeepSeek is leveraging High-Flyers ~50k hopper GPUs (similar in size to the cluster on which OpenAI is believed to be training GPT-5), but what seems likely is that theyre dramatically reducing costs (inference costs for their V2 model, for example, are claimed to be 1/7 that of GPT-4 Turbo). Their subversive (though not new) claim that started to hit the US AI names this week is that more investments do not equal more innovation. Liang: Right now I dont see any new approaches, but big firms do not have a clear upper hand. Big firms have existing customers, but their cash-flow businesses are also their burden, and this makes them vulnerable to disruption at any time. And when asked about the fact that GPT5 has still not been released: OpenAI is not a god, they wont necessarily always be at the forefront.UBSThroughout 2024, the first year we saw massive AI training workload in China, more than 80-90% IDC demand was driven by AI training and concentrated in 1-2 hyperscaler customers, which translated to wholesale hyperscale IDC demand in relatively remote area (as power-consuming AI training is sensitive to utility cost rather than user latency). If AI training and inference cost is significantly lower, we would expect more end users would leverage AI to improve their business or develop new use cases, especially retail customers. Such IDC demand means more focus on location (as user latency is more important than utility cost), and thus greater pricing power for IDC operators that have abundant resources in tier 1 and satellite cities. Meanwhile, a more diversified customer portfolio would also imply greater pricing power.Well update the story as more analysts react.