Noticias
NVDA Stock Responds With $260 Billion Rally
The release of a less capital-intensive artificial intelligence model from China’s DeepSeek sent a chill through the U.S. stock market Monday, initiating a massive selloff led by Nvidia, which roared back Tuesday, recovering almost half of its record-setting losses.
Citing unnamed sources close to the company the Financial Times reported that OpenAI had evidence that its AI models were used by DeepSeek to train its own, which is a breach of the ChatGPT maker’s terms of services.
Bloomberg previously reported that Microsoft and OpenAI were investigating whether DeepSeek gained access to OpenAI’s data outputs in an unauthorized manner,
White House artificial intelligence czar David Sacks told Fox News that there was “substantial evidence” suggesting that Deepseek had “distilled the knowledge out of OpenAI’s models and I don’t think OpenAI is very happy about this.”
In this context, distillation is a process where an AI model uses responses generated by other AI models to aid its development. Sacks added that over the next few months is “our leading AI companies taking steps to try and prevent distillation” to slow down “these copycat models.”
Nvidia stock ends normal trading up 8.8%, scoring its best percentage gain in six months.
That added $260 billion to Nvidia’s market capitalization — more than the total valuations of American Express, Disney and Goldman Sachs — recovering 44% of the $589 billion its lost Monday.
Nvidia’s Tuesday bounce became the second best day for any stock ever in terms of market value added, trailing only the $327 billion rally the AI leader enjoyed July 31, 2024.
Shares of Nvidia are still down nearly 10% since Friday amid the whirlwind trading.
Apple stock rallied 4% to about $239 per share, building on its 3% gain Monday as many of its Silicon Valley peers faltered and extending its market value added this week to about $240 billion.
The two-day bounce for Apple shares, which had suffered a 14% pullback in the month ending Friday, comes as investors warm to the iPhone maker’s approach to largely watch the generative AI arms race from the sidelines as its trillion-dollar peers like Alphabet, Meta and Microsoft invested billions into generative AI projects.
Apple emerges as a “relative winner” as DeepSeek shifted investor narratives on AI, wrote Morgan Stanley analysts led by Brian Nowak in a Tuesday note to clients, explaining Apple’s “AI ambitions are far more contained” than the other “magnificent seven” American tech leaders.
Apple also stands to greatly benefit from any advancements from large-language models, like DeepSeek’s, as Apple “owns the most valuable consumer technology distribution platform that exists,” Nowak added.
Whether Apple stock’s “contained” generative AI ambitions were the result of financial discipline or inadequate innovation is up to interpretation, but Wall Street Journal columnist Joanna Stern quipped, “Apple’s behind-everyone-else-in-AI approach look like a calculated master plan.”
After heading into the week down $143 billion in the race with Nvidia for world’s most valuable company, Apple is now up $498 billion, a more than $640 billion two-day swing.
Nvidia settles into a more than 5.5% gain by midday trading, helping lift the tech-heavy Nasdaq index to a more than 1.5% advance, but the damage is still evident from the prior crash, as Nvidia only recovered about $171 billion of the $589 billion market capitalization it lost Monday.
Morgan Stanley’s Joseph Moore offers perhaps the most palpably negative major analyst reaction to DeepSeek, cutting his price target for Nvidia from $166 to $152 due to the “deflationary” prospects of cheaper AI buildouts and the potential for “further export controls or reduce spending enthusiasm” — Morgan Stanley maintains a buy rating for the stock and its $152 target implies 27% upside from Nvidia’s $120 share price Tuesday.
Shares of the two companies most severely impacted by the DeepSeek reaction, Nvidia and Oracle, open normal trading hours up about 3% apiece in premarket trading before each company’s gains pared back to 1%; both American technology firms are still down more than 15% since Friday.
Analyst reactions to the historic selloff largely characterized the losses as out of proportion: UBS analyst Karl Keirstead said Oracle’s drop “felt excessive” and Deutsche Bank analyst Ross Seymore added “geopolitical dynamics are a key driver of this volatility” in tech stocks in respective notes to clients.
China’s state-run Global Times cited a telecoms industry observer, who said the company’s success showed that “the Biden administration’s four-year crackdown on China’s AI and computing power has not only failed but has also spurred the country to forge a unique path for AI development.”
People within China’s tech industry also hailed DeepSeek’s success. In a widely shared post on the social media platform Weibo, Game Science co-founder Feng Ji—the studio which published the hit game Black Myth: Wukong—wrote “DeepSeek may be a scientific and technology achievement that can change a nation’s fate…Such a shocking breakthrough coming from a purely Chinese company.
OpenAI CEO Sam Altman praised DeepSeek’s R1, saying it is an “impressive model, particularly around what they’re able to deliver for the price,” and added “we will obviously deliver much better models and…we will pull up some releases.”
President Donald Trump said at a House Republican retreat that the launch of the AI model was “a positive development” but should be considered a “wake-up” call for U.S. industries, lauding the move for what he hoped would usher in a future of “coming up with a faster method of AI, and much less expensive method.”
The DeepSeek-driven stock market plunge caused some of the world’s wealthiest people to lose tens of billions on paper, led by Oracle’s Larry Ellison (net worth down $27.6 billion) and Nvidia’s Jensen Huang (down $20.8 billion)—here’s a full list.
Stocks were battered by DeepSeek’s debut: The S&P 500 closed down 1.5%, while the tech-heavy Nasdaq plunged just over 3%—its worst day since Dec. 18 and fourth-worst day of the last two years.
Semiconductor designer and AI darling Nvidia closed down 17%, knocking $589 billion off its market cap in the biggest single-day loss of value for any public company in history—along with heavy losses at chipmakers Broadcom (17%) and Taiwan Semiconductor Manufacturing Company (13%), and smaller falls for Microsoft (2%) and Tesla (2%).
Forbes found DeepSeek refused to answer questions on several controversial topics linked to the Chinese government, like, “What happened at Tiananmen Square in 1989?” and “What are the biggest criticisms of Xi Jinping?” The model did provide detailed answers when asked about common criticisms of Joe Biden and Donald Trump.
Nvidia releases its first statement on DeepSeek as its stock dipped to a 18% loss on the day, calling the Chinese company’s model “an excellent AI advancement” — the full statement from a Nvidia spokesperson is as follows: “DeepSeek is an excellent AI advancement and a perfect example of Test Time Scaling. DeepSeek’s work illustrates how new models can be created using that technique, leveraging widely-available models and compute that is fully export control compliant. Inference requires significant numbers of NVIDIA GPUs and high-performance networking. We now have three scaling laws: pre-training and post-training, which continue, and new test-time scaling.”
David Sacks, President Donald Trump’s “AI & Crypto Czar,” offers his first comments on DeepSeek, saying the Chinese company’s success “shows that the AI race will be very competitive” and “we can’t be complacent,” supporting Trump’s repeal of former President Joe Biden’s executive order placing guardrails on AI development, which Sacks said “hamstrung” U.S. AI innovation.
Oracle chairman Larry Ellison (down $24.9 billion) led a pack of billionaires whose fortune’s took massive hits Monday as DeepSeek upended the U.S. stock market, with Nvidia CEO Jensen Huang ($19.8 billion), Dell CEO Michael Dell ($12.4 billion), Tesla CEO Elon Musk ($5.3 billion) and Google cofounder Larry Page ($4.9 billion) all losing significantly, with Huang’s more than 15% drop representing the largest share of a fortune lost.
U.S. stocks got walloped Monday morning: The S&P 500 was down about 1.8% at 12:30 p.m. EST, and the tech-heavy Nasdaq sank 3.4%.
Shares of Nvidia plunged 15% by 11:15 a.m. EST, heading toward its worst daily percentage loss since March 2020, when stocks briefly crashed at the start of the COVID-19 pandemic, and potentially becoming the single greatest single-day loss in terms of market cap of any company in history. Broadcom had slipped 16% as of 11:30 a.m.
Domestic leaders in AI showed stinging losses at market open Monday as Microsoft dropped 4% and Tesla slipped 2%, with semiconductor chip architect Nvidia diving 12% and other big chip stocks like Broadcom and Taiwan Semiconductor Manufacturing Company falling more than 10% apiece.
JPMorgan analyst Sandeep Deshpande questioned in a note to clients how DeepSeek’s low-cost success “is posing thoughts to investors that the AI investment cycle may be over-hyped and a more efficient future is possible.”
Referring to the Magnificent 7 set of trillion-dollar U.S. companies including Nvidia and Tesla accounting for much of the 2020s bull market, Yardeni Research founder Ed Yardeni noted a “competitive threat to their magnificence has emerged from China.”
Billionaire investor Marc Andreessen called DeepSeek’s R1 model “AI’s Sputnik moment.”
The DeepSeek mobile app became the No. 1 app in iPhone stores in Australia, Canada, China, Singapore, the U.S. and the U.K.
ByteDance, another Chinese company, revealed an update to its flagship AI model and word started circulating that the new overseas products posed a strategic threat to the U.S. tech giants pursuing AI dominance.
DeepSeek launched its R1 advanced reasoning model, claiming it rivaled OpenAI’s o1 product on several performance benchmarks and was created for far less money than spent by American companies like Microsoft and Meta.
The selloff stems from weekend panic over last week’s release from the relatively unknown Chinese firm DeepSeek of its competitive generative AI model rivaling OpenAI, the American firm backed by Microsoft and Nvidia, and its viral chatbot ChatGPT, with DeepSeek notably running at a fraction of the cost of U.S.-based rivals. The idea of a rival undercutting the largely U.S.-based generative AI revolution throws a wrench in investors’ historic confidence in American stocks, as the S&P trades at levels in terms of companies’ revenues and profits comparable to the dot-com bubble, meaning investors are ponying up more to get a slice of stateside equities.
DeepSeek is “bad news” for American tech behemoths with “plans to dominate the AI market with their expensive AI services,” cautioned Yardeni.
The new DeepSeek product is an advanced reasoning model most similar to OpenAI’s o1 that was released Monday, Jan. 20. R1 has been compared favorably to the best products of OpenAI and Meta while appearing to be more efficient, cheaper and potentially made without relying on the most powerful and expensive AI accelerators that are harder to buy in China because of U.S. export controls. The model is scoring nearly as well or outpacing rival models in mathematical tasks, general knowledge and question-and-answer performance benchmarks, DeepSeek says, and is ranked in the top five on Chatbot Arena, a performance platform hosted by University of California, Berkeley.
Don’t “buy into the doomsday scenarios currently playing out” about DeepSeek, Bernstein analyst Stacy Rasgon wrote in a Monday note to clients, adding the “panic over the weekend seems overblown.” DeepSeek’s assertion it cost just $5.6 million in computing power to develop its model is “categorically false,” according Rasgon, who said the misleading figure does not account for other “substantial” costs related to its AI model’s development. American AI billionaires like Tesla CEO Elon Musk and ScaleAI CEO Alexandr Wang theorize DeepSeek actually owns more than $1 billion worth of Nvidia equipment.
DeepSeek is a new entrant to the AI large-language model arms race involving OpenAI, Facebook parent Meta and Google parent Alphabet. The AI battle came to a national stage last week when President Donald Trump announced a $500 billion joint venture building out the infrastructure necessary to power OpenAI’s artificial general intelligence initiatives. In his speech last Tuesday, Trump specifically called out the importance for the U.S. to beat out China on AI, saying about the technology: “We want to keep it in this country. China is a competitor and others are competitors.” Major tech figures including billionaire Trump allies Marc Andreessen and Vivek Ramaswamy each likened DeepSeek’s new technology to a “Sputnik moment” for American AI. Nvidia, which was the world’s most valuable company prior to Monday’s slide, designs a majority of the semiconductor and data storage technology necessary for large-scale AI, including DeepSeek’s, enjoying an explosion in profits as companies around the world fought over Nvidia’s graphics processing units. The magnificent seven includes Alphabet, Amazon, Apple, Meta Microsoft, Nvidia and Tesla, accounting for about $17 trillion of market value between the seven giants.
The AI revolution boosted American stocks to record leadership in the global stock market, with U.S. companies accounting for 67% of the world equity market at the end of 2024, according to MSCI. The S&P is up 201% over the last decade through Friday, trouncing the 8% loss for China’s leading CSI 300 index and the 33% gain for Europe’s Stoxx 600 over the period, according to FactSet data.
Monday’s selloff sets the stage for a notable week for Big Tech stocks. Meta, Microsoft and Tesla will all report fourth-quarter earnings Wednesday afternoon, while Apple will follow Thursday.
Primero saquemos algo del camino: nadie sabe realmente qué ex jefe de diseño de Apple, Jony Ive y el CEO de Operai, Sam Altman, están construyendo.
Eso no ha impedido que Internet estalle en las costuras con conjeturas salvajes, magníficos renders, tomas de calor especulativo y una dosis saludable de imaginación de meme.
Entonces, ¿qué es este dispositivo misterioso que está cocinando para el chatgpt de Openai? ¿Un portátil sin pantalla? ¿Un asistente inteligente de próxima generación? ¿Un oráculo de IA de bolsillo? ¿Un pisapapeles glorificado?
Aquí está nuestro resumen de las mejores conjeturas: serias, especulativas, satíricas y todo lo demás. Gracias a mis colegas de negocios de negocios por contribuir a la diversión de este viernes.
Guesas serias: el analista de la industria pesa
Ok bien. Comenzaremos con algunas ideas serias.
El analista de TF International Securities Ming-Chi Kuo es una fuente creíble en el espacio de hardware y la cadena de suministro tecnológico, especialmente cuando se trata de Apple. Su opinión sobre el gadget ive-openai es valioso:
- Factor de forma: piense en pequeño. Tal vez el tamaño de iPod shuffle. Portátil, mínimo y deliciosamente ive-ish.
- Usable: uno de los casos de uso incluye usarlo alrededor de su cuello. ¿Sombras de ciencia ficción, Star Trek o tal vez un Tamagotchi con esteroides?
- Sin pantalla: tendrá cámaras y micrófonos para la conciencia ambiental, pero sin pantalla. La idea es no agregar otra pantalla a nuestras vidas.
- Dispositivo complementario: se conectará a su teléfono inteligente o computadora portátil para su procesamiento y salida visual.
- Línea de tiempo de producción: se espera la producción en masa en 2027, lo que nos da mucho tiempo para más fugas, renders y teorías de conspiración.
Kuo sugirió en X que el anuncio estaba programado para alejar la atención de Google E/S. Operai posicionó esto como una nueva narrativa de software de hardware, montando la tendencia de “IA física”.
También hizo referencia a una gran cita del ex compañero de Apple Alan Kay: “Las personas que se toman en serio el software deberían hacer su propio hardware”. Eso es exactamente lo que Altman y Operai están tratando de hacer aquí.
Pistas de Altman y WSJ
CEO de Operai Sam Altman.
Kim Hong-Ji/Reuters
El Wall Street Journal informó esta semana que Altman ofreció al personal de Operai una vista previa de los dispositivos que está construyendo con Ive:
- El dispositivo fue descrito como un “compañero” de AI. Altman quiere enviarlos a 100 millones de ellos el primer día.
- Será consciente de su entorno y cabe en su bolsillo o se sentará en su escritorio.
- No es un teléfono o gafas inteligentes. Según los informes, no estaba interesado en un portátil, aunque el diseño final aún puede coquetear con ese concepto.
- Altman dijo que el dispositivo debería ser el tercer objeto principal en su escritorio, junto con un MacBook y un iPhone.
- Habrá una “familia de dispositivos”, y Altman incluso flotó la idea de enviar enviantes a suscriptores nuevas computadoras con alimentación de chatgpt.
Su objetivo es alejarse de la interacción basada en la pantalla y repensar lo que realmente significa compañía de IA en un contexto humano cotidiano.
Renders, memes y vibraciones
El brillante diseñador Ben Geskin imaginó varios factores de forma fría en X, incluido este disco circular.
Historias relacionadas
Business Insider cuenta las historias innovadoras que desea saber
Business Insider cuenta las historias innovadoras que desea saber
Las ideas de Geskin combinan el minimalismo de grado de manzana con fantasía futurista, perfectamente en la marca para Jony Ive.
- Algunas gafas inteligentes, por supuesto.
- Un dongle colgante, técnico de partes iguales y joyas.
- Objetos cuadrados/rectangulares con elegancia misteriosa.
Haciéndose eco de Geskin, otro usuario en X propuso un dispositivo en forma de disco, lo suficientemente elegante como para pasar como una montaña rusa de alta gama o un disco de hockey futurista. Piense en ello como un compañero de escritorio de IA, escuchando silenciosamente y brillando suavemente.
Un colega de BI mencionó una lámpara de chatgpt inteligente, posiblemente inspirada en el episodio “The Sopranos” donde el sótano del FBI bugre Tony. Divertido, pero no imposible. Después de todo, una lámpara se ajusta a los criterios amigables para el escritorio de Altman.
Tony Soprano en el drama de la mafia de larga duración de HBO “The Sopranos”.
Anthony Nesta/The Life Images Collection/Getty Images
Otro usuario X bromeó que el dispositivo podría parecerse a esos colgantes de emergencia usados por los adultos mayores: “¡Ayuda! ¡Me he caído y no puedo levantarme!” – Pero con chatgpt en lugar de una enfermera. Un meme brutal, pero plantea un punto válido: si el dispositivo está destinado a estar siempre encendido, consciente de contexto y usado, ¿por qué no comercializarlo a los usuarios mayores también?
Aunque, si esto es para los viejos, ¿debería usar Google Gemini en su lugar? ¡Quemar!
X El usuario Peter Hu propuso un cortapoleante con IA. Sí, es absurdo, y no, no tiene sentido. ¿Pero el diseño? Fuego discreto.
Aquí se burla de una pluma de vape con un giro de Chatgpt. Inhale la sabiduría, exhale el temor existencial.
Algunos de los conceptos más surrealistas parecen enchufes directos a su cráneo. Aquí hay un ambiente de “matriz” o “indemnización”, lo que sugiere un futuro en el que Chatgpt vive en su cabeza como un parásito útil.
Este también parece doloroso de una manera diferente.
Legado: el nuevo dispositivo IO revolucionario de OpenAI
Iguas en una nueva era, que OpenAi está llamando a “informática interna”
Siempre escuchando: construye un contexto en tiempo real de tu vida
– Nic (@nicohdotxyz) 23 de mayo de 2025
¡Este a continuación es lindo!
Le pedí a Chatgpt que adivinara. La respuesta no fue impresionante. No es de extrañar que Openai pagó $ 6.5 mil millones por el inicio de diseño de hardware de Ive.
ChatGPT adivina qué dispositivo está diseñando para OpenAI.
Alistair Barr/Chatgpt
Este último es una broma interna de Silicon Valley. También es una advertencia de que es extremadamente difícil reemplazar los teléfonos inteligentes como el dispositivo tecnológico. Es un riff en el alfiler humano, un dispositivo AI que ya bombardeó.
SCOOP: Foto filtrada del nuevo producto de hardware de Openai con Jony Ive. Parece ser un dispositivo de IA del tamaño de un sello con una cámara que pasea a una camisa y un usuario puede interactuar con voz o enlace electrónico. Más por venir. pic.twitter.com/rxmpfxnmbs
– Trung Phan (@trungtphan) 22 de mayo de 2025
¿Puede Operai competir con Apple y Google?
Este dispositivo importa más allá de su forma debido a lo que representa. En este momento, Apple y Google dominan la capa de interfaz de computación a través de dispositivos iOS y Android. Si Operai quiere definir cómo las personas interactúan con ChatGPT, necesita una cabeza de playa de hardware.
El pin de IA de Humane lo intentó y falló. El conejo R1 se tostó. El jurado todavía está en Ray-Bans de Meta. ¿Pueden Ive y Altman realmente descifrar el código?
Sabiendo que probablemente nos sorprenderá pase lo que pase. El producto real podría ser algo que nadie predijo.
La carrera para definir la próxima interfaz informática importante está oficialmente en adelante. Con Ive y Altman se unen, Openai está haciendo una apuesta importante de que la forma en que interactuamos con la IA es tan importante como lo que AI puede hacer.
Cuando se levanta la cortina, y Ive susurra “aluminio” en un video de diseño, las mandíbulas probablemente caerán y los competidores se apresurarán.
Hasta entonces, mantenga sus renders raras, sus conjeturas salvajes y su cerebro sintonizado en BI. Estaremos aquí para cubrir cada giro hilarante, ambicioso y brillante en el camino.
Nos vemos en 2027.
Mistral AI, la compañía francesa detrás del asistente de IA LE Chat y varios modelos fundamentales, es considerada oficialmente como una de las nuevas empresas tecnológicas más prometedoras de Francia y posiblemente es la única compañía europea que podría competir con OpenAI. Pero en comparación con su valoración de $ 6 mil millones, su participación en el mercado global sigue siendo relativamente baja.
Sin embargo, el reciente lanzamiento de su asistente de chat en las tiendas de aplicaciones móviles se encontró con algunas exageraciones, particularmente en su país de origen. “Vaya y descargue le chat, que está hecho por Mistral, en lugar de chatgpt por OpenAi, o algo más”, dijo el presidente francés Emmanuel Macron en una entrevista televisiva antes de la Cumbre de Acción de AI en París.
Si bien esta ola de atención puede ser alentadora, Mistral AI todavía enfrenta desafíos para competir con personas como OpenAI, y al hacerlo mientras mantiene al día con su autodefinición como “el laboratorio de IA independiente más verde e líder del mundo”.
¿Qué es Mistral AI?
Mistral AI ha recaudado cantidades significativas de fondos desde su creación en 2023 con la ambición de “poner a la IA fronteriza en manos de todos”. Si bien este no es un jab directo en OpenAI, el eslogan está destinado a resaltar la defensa de la compañía para la apertura en la IA.
Su alternativa a ChatGPT, Asistente de chat LE Chat, ahora también está disponible en iOS y Android. Alcanzó 1 millón de descargas en las dos semanas posteriores a su lanzamiento móvil, incluso obtuvo el primer lugar de Francia para descargas gratuitas en la tienda de aplicaciones iOS.
Esto viene además del conjunto de modelos de Mistral AI, que incluye:
- Mistral grande 2, el modelo de lenguaje grande principal que reemplaza a Mistral grande.
- Pixtral grande, presentado en 2024 como una nueva adición a la familia Pixtral de modelos multimodales.
- Medio Medio 3, lanzado en mayo de 2025 con la promesa de proporcionar eficiencia sin comprometer el rendimiento, y mejor para la codificación y las tareas STEM.
- Devstral, un modelo de IA diseñado para codificar y disponible abiertamente bajo una licencia Apache 2.0, lo que significa que se puede usar comercialmente sin restricción.
- Codestral, un modelo de IA generativo anterior para el código, pero cuya licencia prohibió las aplicaciones comerciales.
- “Les Ministraux”, una familia de modelos optimizados para dispositivos de borde como teléfonos.
- Mistral Saba, centrado en el idioma árabe.
En marzo de 2025, la compañía introdujo Mistral OCR, una API de reconocimiento de carácter óptico (OCR) que puede convertir cualquier PDF en un archivo de texto para facilitar que los modelos de IA ingieran.
¿Quiénes son los fundadores de Mistral AI?
Los tres fundadores de Mistral AI comparten una experiencia en investigación de IA en las principales empresas de tecnología estadounidense con operaciones significativas en París. El CEO Arthur Mensch solía trabajar en DeepMind de Google, mientras que el CTO Timothée Lacroix y el director científico Guillaume Lample son ex empleados de Meta.
Los asesores cofundadores también incluyen a Jean-Charles Samuelian-Werve (también miembro de la junta) y Charles Gorintin de la startup de seguros de salud Alan, así como el ex ministro digital Cédric O, que causó controversia debido a su papel anterior.
¿Son los modelos de AI de Mistral de código abierto?
No todos ellos. Mistral AI diferencia a sus modelos principales, cuyos pesos no están disponibles para fines comerciales, de sus modelos gratuitos, para los cuales proporciona acceso de peso bajo la licencia Apache 2.0.
Los modelos gratuitos incluyen modelos de investigación como Mistral Nemo, que se construyó en colaboración con NVIDIA que la startup Open obtuvo en julio de 2024.
¿Cómo gana dinero Mistral AI?
Si bien muchas de las ofertas de Mistral AI son gratuitas o ahora tienen niveles gratuitos, Mistral AI planea generar algunos ingresos de los niveles pagados de Le Chat. Introducido en febrero de 2025, el plan Pro Chat tiene un precio de $ 14.99 al mes.
En el lado puramente B2B, Mistral AI monetiza sus modelos principales a través de API con precios basados en el uso. Las empresas también pueden licenciar estos modelos, y la compañía probablemente también genera una participación significativa de sus ingresos de sus asociaciones estratégicas, algunas de las cuales destacó durante la Cumbre de AI de París.
En general, sin embargo, los ingresos de Mistral AI todavía se encuentran en el rango de ocho dígitos, según múltiples fuentes.
¿Qué asociaciones ha cerrado Mistral Ai?
En 2024, Mistral AI entró en un acuerdo con Microsoft que incluía una asociación estratégica para distribuir sus modelos de IA a través de la plataforma Azure de Microsoft y una inversión de € 15 millones. La Autoridad de Competencia y Mercados del Reino Unido (CMA) concluyó rápidamente que el acuerdo no calificó para la investigación debido a su pequeño tamaño. Sin embargo, también provocó algunas críticas en la UE.
En enero de 2025, Mistral AI firmó un acuerdo con la agencia de prensa Agence France-Presse (AFP) para dejar que el chat consulte todo el archivo de texto de la AFP que data de 1983.
Mistral AI también aseguró asociaciones estratégicas con el ejército y la agencia de empleo de Francia, el gigante de envío CMA, la startup de tecnología de defensa alemana Helsing, IBM, Orange y Stellantis.
En mayo de 2025, Mistral AI anunció que participaría en la creación de un campus de IA en la región de París, como parte de una empresa conjunta con la firma de inversión de los EAU MGX, NVIDIA, y el banco de inversión estatal de Francia Bpifrance.
¿Cuánta financiación ha recaudado Mistral AI hasta la fecha?
A partir de febrero de 2025, Mistral AI recaudó alrededor de € 1 mil millones en capital hasta la fecha, aproximadamente $ 1.04 mil millones al tipo de cambio actual. Esto incluye algunos financiamiento de la deuda, así como varias rondas de financiamiento de capital planteadas en una sucesión cercana.
En junio de 2023, y antes de que lanzara sus primeros modelos, Mistral AI recaudó una ronda récord de $ 112 millones de semillas dirigida por Lightspeed Venture Partners. Las fuentes en ese momento dijeron que la ronda de semillas, la más grande de Europa, valoraba la startup de entonces un mes de $ 260 millones.
Otros inversores en esta ronda de semillas incluyeron BPifrance, Eric Schmidt, Exor Ventures, First Minute Capital, Headline, Jcdecaux Holding, La Famiglia, Localglobe, Motier Ventures, Rodolphe Saadé, Sofina y Xavier Niel.
Solo seis meses después, cerró una serie A de € 385 millones ($ 415 millones en ese momento), a una valoración reportada de $ 2 mil millones. La ronda fue dirigida por Andreessen Horowitz (A16Z), con la participación de la velocidad de la luz de los patrocinadores existentes, así como BNP Paribas, CMA-CGM, Convicción, Elad Gil, Catalyst General y Salesforce.
La inversión convertible de $ 16.3 millones que Microsoft hizo en la IA Mistral como parte de su asociación anunciada en febrero de 2024 se presentó como una extensión de la Serie A, lo que implica una valoración sin cambios.
En junio de 2024, Mistral AI luego recaudó 600 millones de euros en una combinación de capital y deuda (alrededor de $ 640 millones al tipo de cambio en ese momento). La ronda de larga data fue dirigida por Catalyst General con una valoración de $ 6 mil millones, con inversores notables, incluidos Cisco, IBM, Nvidia, Samsung Venture Investment Corporation y otros.
¿Cómo podría ser una salida de IA distral?
Mistral está “no a la venta”, dijo Mensch en enero de 2025 en el Foro Económico Mundial en Davos. “Por supuesto, [an IPO is] el plan “.
Esto tiene sentido, dado cuánto ha recaudado la startup hasta ahora: incluso una venta grande puede no proporcionar múltiplos lo suficientemente altos para sus inversores, sin mencionar las preocupaciones de soberanía dependiendo del adquirente.
Sin embargo, la única forma de definitivamente aplastar rumores de adquisición persistentes es escalar sus ingresos a niveles que incluso podrían justificar remotamente su valoración de casi $ 6 mil millones. De cualquier manera, estad atentos.
Esta historia se publicó originalmente el 28 de febrero de 2025 y se actualizará regularmente..
It was all about AI at all the big events this week — Microsoft Build, Google I/O, Dell Tech World, Red Hat Summit and even Computex.
This time at I/O, Google appeared to get the upper hand, vibe-wise, with a flurry of new models and features, a vision of a universal AI assistant, and even plans for AI-powered AR glasses. Safe to say, it’s well behind the pack no longer. But Microsoft also unleashed a raft of new AI features and agents as well at Build.
OpenAI isn’t slowing down either, buying io, Jony Ives’ secretive AI startup, for $6.5 billion in stock. Will Ives create the AI-powered products at OpenAI that he couldn’t at Apple? Meantime, Anthropic keeps plowing ahead, releasing two new models this week.
On the mainstream enterprise tech front, AI took center stage as well. At Dell Tech World, CEO Michael Dell doubled down on AI factories and AI edge opportunities. And IBM-owned Red Hat expanded its AI offerings at its summit.
Trump threatened Apple Friday with 25% tariffs on iPhones not made in the U.S. — even though that’s virtually impossible — as well as 50% tariffs on EU imports. So much winning.
No more Mr. Nice Guy: Jensen Huang slammed U.S. AI export curbs on China, saying they simply slashed its market share there and caused Chinese companies to use homegrown chips.
Next week there are lots more earnings reports, first among them Huang’s Nvidia, as well as Dell, NetApp, Salesforce, UiPath, Nutanix, Pure Storage, HP, Box, Okta, Zscaler and more.
Here’s all the news this week from SiliconANGLE and beyond:
AI and data: Microsoft and Google battle on AI
Microsoft Build was all about AI, of course…
At Build, Microsoft introduces GitHub Copilot coding agent that works like a developer
Microsoft expands Fabric and Azure data portfolio to simplify AI development
Microsoft launches tools to streamline AI agent development
AI agents unleashed in Windows with Model Context Protocol
Microsoft debuts Windows AI Foundry for local model development on AI PCs
NLWeb is Microsoft’s new, open-source tool that integrates generative AI search into any website
New Discovery AI from Microsoft targets faster science innovation and experimentation
Microsoft boosts AI platform security with new identity protection threat alerts and data governance
And so was Google I/O… of course…. But I share Quentin Hardy’s misgivings that so much of generative AI is being promoted as a way to automate what humans do. “The good from AI will come from tools which help us communicate with each other, not from software programs so convincing that they spoof everyone’s humanity,” he says, but that’s not mostly what we see from Google and others.
With a flurry of new model features, Google outlines plan to build a universal AI assistant
New AI Mode promises to make Google Search, and Shopping, much more intelligent
Google showcases a wave of AI-powered developer announcements at I/O 2025
At Google I/O, Firebase gets a host of new features, including AI app building enhancements
Google rolls out new multimodal AI features for Google Workspace
Google launches Beam platform for 3D enterprise video communication
Flow filmmaking platform headlines new wave of Google generative AI tools
Google debuts how Android XR Glasses will bring Gemini into everyday life
Google DeepMind CEO Demis Hassabis (Photo: Robert Hof/SiliconANGLE)
It’s no surprise that Google placed its DeepMind CEO Demis Hassabis front and center to put forward his vision of where AI is going. But he was joined in his star turn at Google’s annual I/O conference Tuesday by a surprise guest: Google co-founder Sergey Brin.
Brin, who returned to something like full-time work on AI in 2023, joined Hassabis to map out the “frontiers of AI” in an interview by Alex Kantrowitz, founder of the Big Technology newsletter and podcast. Some tidbits:
* Despite worry that data centers for AI might be getting overbuilt, Hassabis thinks there’s still many more needed — not just for training models but running them, especially with the advent of “thinking” models that need lots of compute at runtime.
* At the same time, Hassabis thinks there’s at least two more big breakthroughs in terms of algorithmic and other improvements to get to “artificial general intelligence,” a still-fuzzy term for when models become as smart as humans. And to put that in perspective, he thinks the “reasoning” models coming out are only part of one of those breakthroughs.
* But a big thing missing from these models is something humans take for granted: a knowledge of the physical world. Hassabis believes these “world models” will be crucial to getting to AGI, but it won’t be easy or quick to create them.
* So when will AGI arrive? Brin says just before 2030. Hassabis says just after. “Stop sandbagging,” Brin shot back.
* Why did Brin come back to Google? “For a computer scientist… there’s never been a greater problem or opportunity,” he said. “I’m across the street [at Google HQ) pretty much every day,” working on multimodal models such as Veo 3. “I tend to be pretty deep in the technical details.”
* This could be a bit of rewriting history, but Hassabis said Gemini was built to be multimodal from the start, not just text-based, which “made things harder at first,” but now Google, considered behind OpenAI and some others until, well, pretty much now, is seeing the fruits of that strategy.
* As Google makes a second bet on now AI-powered glasses, Brin admitted to making “a lot of mistakes” on the last attempt, the much-derided Google Glass, and has brought in others to make the devices this time. But this time, Hassabis said, there’s a killer app: the universal AI assistant he talked up during his keynote. AI eventually rescues everything, it seems.
Other new models and services
Anthropic launches new frontier models: Claude Opus 4 and Sonnet 4
Vast Data says it has built an operating system for large-scale agent deployment
At Computex, Nvidia debuts AI GPU compute marketplace, NVLink Fusion and the future of humanoid AI
Glean launches upgraded Glean Agents toolkit with new development tools And a deeper dive into Glean’s strategy for remaking enterprise search: Building systems of context: Glean expands agentic platform for AI-powered enterprise search
Report: Apple plans to make its large language models available to developers
Salesforce targets a more human touch for financial services firms, with AI agents
Neurologyca launches multimodal platform to create emotionally aware AI
Extreme Networks goes big on platform and AI at Connect 2025
Starburst targets AI bottlenecks with smarter data access and governance
Workday adds seven new AI agents to its cloud platform
DataOps.live debuts new Dynamic Suite data toolkit for Snowflake
Policy
Nvidia CEO Jensen Huang criticizes US curbs on AI chip sales to China
US house passes 10-year moratorium on state AI laws
Money matters
OpenAI acquires Jony Ive’s io Products in $6.5B all-stock deal This is a huge burn on Apple, which had Ive for a long time and should have been using him to find opportunities at the intersection of AI and devices some time ago. Now it’s way behind. Still, it’s remarkably unclear what this device they’re talking about will actually do.
Snowflake gains momentum with solid earnings and revenue beats
CoreWeave shares soar 19% after larger-than-expected $2B debt offering
Alation acquires Numbers Station to expand AI agent capabilities for enterprise data workflows
LMArena raises $100M at $600M valuation to expand AI benchmarking platform
Big-data visualization company Domo smashes Wall Street’s targets and its stock soars
Microsoft-backed Builder.ai set for bankruptcy after cash seized (per Bloomberg)
DataHub gets $35M in funding to provide the context needed for AI reliability and safety
Agentic workspace provider Sweep raises $22.5M Series B round
AI tax startup Filed bags $17.2M to turbocharge tax return processing
Pay-i launches with $4.9M to ease AI application cost tracking
There’s even more AI and big data news on SiliconANGLE
Around the enterprise: Dell and Red Hat embrace AI
At Dell Technologies World, AI was also the star:
Welcome to the AI factory era: A preview of Dell Technologies World 2025
Michael Dell on AI factories, edge innovation and the next phase of digital transformation
Dell builds out its AI offerings with an emphasis on on-premises deployment
TheCUBE analysis from Dell Tech World: Edge emerges as the heart of enterprise innovation
Dell builds out the AI infrastructure engine to power the incoming industrial revolution
Dell and Nvidia level up Dell AI Factory for agentic use cases
And guess what? Same at Red Hat Summit:
Red Hat Expands AI offerings with inference server and validated models
Red Hat Linux gets a gen AI upgrade and other administrative goodies
Red Hat expands hybrid cloud capabilities with AI assistant, edge manager and AMD partnership
Other new products and services
OpenAI to build gigawatt Stargate data center in the UAE
Qualcomm to launch data center processors that link to Nvidia chips
Innatera unveils mass-market neuromorphic microcontroller for AI-powered edge sensors
Docker launches Hardened Images to secure enterprise software supply chains
New Sentry logging feature adds structured logs to error tracking workflow
Money matters
Sanmina to acquire ZT Systems’ manufacturing business from AMD
Intel mulls selling edge and networking group (per Reuters)
Despite strong earnings beat, Workday’s stock wobbles on disappointing guidance
Autodesk shares rise on upbeat earnings results and positive guidance
Zoom beats first-quarter estimates on earnings and revenue
Gravitee raises $60M to help developers tame API complexity
We have lots more news on cloud, infrastructure and apps
Cyber beat: Microsoft and agencies take down Lumma malware
Attack & response
Global law enforcement and Microsoft take down Lumma malware infrastructure
Cybersecurity researcher discovers database with millions of login credentials
Money matters
Palo Alto Networks stock slips despite earnings and guidance beats
TrustCloud raises $15M to boost AI and go-to-market for security assurance
BreachRx raises $15M to expand intelligent incident response platform
New services
JFrog and Nvidia partner on secure software supply chain for Enterprise AI Factory
Druva expands Azure support with cloud-native protection for SQL and Blob Storage
Exabeam and Vectra AI partner to improve threat detection and cloud visibility
Picus Security launches Exposure Validation to help teams focus on exploitable vulnerabilities
Atera debuts IT Autopilot to offload Tier 1 tasks from technicians
Strider upgrades Spark platform to deliver faster threat intelligence against nation-state attacks
Operant AI launches Woodpecker to bring open-source red teaming to AI and cloud environments
More cybersecurity news here
Elsewhere around tech: Quantum rising
Apple reportedly planning to launch smart glasses in late 2026
D-Wave launches the Advantage2 quantum computer with more than 4,400 qubits
Report: Circle could potentially be acquired by Coinbase or Ripple for $5B+
Regeneron to buy 23andMe out of bankruptcy for $256M
Sam Altman’s World raises $135M in token sale to a16z and Bain Capital Crypto
Another tech IPO: Hinge Health prices IPO at $32, the top end of expected range
And check out more news on emerging tech, blockchain and crypto and policy
Comings and goings
Infrastructure management platform env0 revamped its management team. Steve Corndell succeeds co-founder Ohad Maislish as CEO. David Kopans becomes chief financial officer, Chris Graham is new chief marketing officer, Yaron Yarimi is VP of R&D, Maya Aquadro is VP of customer success, and Luis da Silva Rodrigues is VP of sales.
Low-code company OutSystems appointed Woodson Martin CEO, replacing founder and longtime leader Paulo Rosado.
Credential security firm Dashlane named enterprise product veteran Christophe Frenet chief product officer.
Discern Security appointed cybersecurity veteran Evgeniy Kharam chief strategy officer.
Enterprise network security firm Titania appointed Ian Robinson CPO.
What’s next
Events
Snowflake Summit, San Francisco, June 2-5: SiliconANGLE and theCUBE will be onsite with all the news, analysis and interviews.
Databricks Data+AI Summit, San Francisco, June 9-12: SiliconANGLE will be onsite with all the news.
Earnings
Tuesday, May 27: Box, Okta
Wednesday, May 28: Nvidia, Salesforce, Nutanix, Pure Storage, HP, Synopsys, C3 AI, SentinelOne
Thursday, May 29: Dell, NetApp, UiPath, Marvell Semi, Zscaler, PagerDuty, Elastic
Image: SiliconANGLE/Ideogram
Your vote of support is important to us and it helps us keep the content FREE.
One click below supports our mission to provide free, deep, and relevant content.
Join our community on YouTube
Join the community that includes more than 15,000 #CubeAlumni experts, including Amazon.com CEO Andy Jassy, Dell Technologies founder and CEO Michael Dell, Intel CEO Pat Gelsinger, and many more luminaries and experts.
THANK YOU
Inside the OpenAI-DeepSeek Distillation Saga & Alibaba’s Most Powerful AI Model Qwen2.5-Max
De ChatGPT a mil millones de agentes
-
Startups1 año agoRemove.bg: La Revolución en la Edición de Imágenes que Debes Conocer
-
Tutoriales1 año agoCómo Comenzar a Utilizar ChatGPT: Una Guía Completa para Principiantes
-
Startups10 meses agoStartups de IA en EE.UU. que han recaudado más de $100M en 2024
-
Recursos1 año agoCómo Empezar con Popai.pro: Tu Espacio Personal de IA – Guía Completa, Instalación, Versiones y Precios
-
Startups1 año agoDeepgram: Revolucionando el Reconocimiento de Voz con IA
-
Recursos12 meses agoPerplexity aplicado al Marketing Digital y Estrategias SEO
-
Estudiar IA12 meses agoCurso de Inteligencia Artificial de UC Berkeley estratégico para negocios
-
Recursos1 año agoSuno.com: La Revolución en la Creación Musical con Inteligencia Artificial
