Elon Musk’s SpaceXAI has been bleeding staff since its merger | TechCrunch
Elon Musk’s newly rebranded SpaceXAI is reportedly losing top talent, with more than 50 researchers and engineers departing since February, according to The Information. The exits include key leaders across coding, world models, and Grok voice.
Rivals like Meta and Thinking Machine Labs are reportedly scooping up former staff, with the company’s core pre-training team dwindling to just a handful of people. Since February, at least 11 xAI employees have defected to Meta, according to The Information’s report. At least seven have left to join Mira Murati’s Thinking Machine Labs. TechCrunch has previously reported on 11 of the xAI departures announced directly after the merger, including two co-founders.
SpaceX acquired xAI — two companies owned by Musk — in February and has since installed new leadership at the company. Musk renamed the combined company SpaceXAI earlier this month.
The pre-training departures, which followed the exit of team lead Juntang Zhuang, have particularly concerned employees and people close to SpaceXAI, per The Information. Pre-training is the first step to building new AI models, and many have questioned whether the company is still committed to developing leading models.
The report also found that Musk’s culture of extreme work led some staff to leave — something Musk employees across his companies, including Tesla, have complained about. A source who spoke to The Information said Musk set unrealistic deadlines for training models, which led to cutting corners on Grok.
Of course, several of the exits could have been driven by a desire to cash out.
SpaceX regularly offers tenders so employees can sell vested shares privately. Others might simply feel confident that their equity is close to liquidity given the company’s blockbuster IPO expectations. Once employees see the financial upside light at the end of the tunnel, they’re less likely to work at a company that puts undue pressure on them and may not be building the leading models they want to work on.
TechCrunch has reached out to SpaceX for comment.
Elon Musk’s newly rebranded SpaceXAI is reportedly losing top talent, with more than 50 researchers and engineers departing since February, according to The Information. The exits include key leaders across coding, world models, and Grok voice.
Rivals like Meta and Thinking Machine Labs are reportedly scooping up former staff, with the company’s core pre-training team dwindling to just a handful of people. Since February, at least 11 xAI employees have defected to Meta, according to The Information’s report. At least seven have left to join Mira Murati’s Thinking Machine Labs. TechCrunch has previously reported on 11 of the xAI departures announced directly after the merger, including two co-founders.
SpaceX acquired xAI — two companies owned by Musk — in February and has since installed new leadership at the company. Musk renamed the combined company SpaceXAI earlier this month.
The pre-training departures, which followed the exit of team lead Juntang Zhuang, have particularly concerned employees and people close to SpaceXAI, per The Information. Pre-training is the first step to building new AI models, and many have questioned whether the company is still committed to developing leading models.
The report also found that Musk’s culture of extreme work led some staff to leave — something Musk employees across his companies, including Tesla, have complained about. A source who spoke to The Information said Musk set unrealistic deadlines for training models, which led to cutting corners on Grok.
Of course, several of the exits could have been driven by a desire to cash out.
SpaceX regularly offers tenders so employees can sell vested shares privately. Others might simply feel confident that their equity is close to liquidity given the company’s blockbuster IPO expectations. Once employees see the financial upside light at the end of the tunnel, they’re less likely to work at a company that puts undue pressure on them and may not be building the leading models they want to work on.
TechCrunch has reached out to SpaceX for comment.
When you purchase through links in our articles, we may earn a small commission. This doesn’t affect our editorial independence.
#Elon #Musks #SpaceXAI #bleeding #staff #merger #TechCrunchElon Musk,SpaceX,spacexai,xAI
![IBM Crosses One of Computing’s Biggest Barriers With World’s First Sub-1 Nanometer Chip
In a major breakthrough, IBM revealed the world’s first semiconductor chip technology built on a sub-1 nanometer chipmaking process. For comparison, the process uses transistor features smaller than the width of a DNA strand, which measures about 2.5 nanometers across. The chip itself is about the size of a fingernail but holds almost 100 billion transistors, and the company expects it could enter markets as early as the next five years. In a statement released today, IBM said the new chip features nearly twice the density of its 2-nanometer chip, released in 2021. According to an accompanying technical report, the chip also demonstrated up to 70% greater energy efficiency than its predecessor. In designing the chip, researchers developed an “entirely new transistor architecture” called nanostack, which “vertically stacks and staggers transistors” to enable IBM’s 0.7-nanometer chip technology, IBM explained. A section of the chip seen with a transmission electron microscope. Credit: IBM “With our new nanostack architecture, we’re not just making smaller transistors,” Jay Gambetta, director of IBM Research, said in the statement. “We’re reinventing how chips are built to deliver dramatically more power and energy efficiency.”
Smaller and smaller Semiconductor chips enable things like computers, home appliances, communications, and transportation devices. In 1965, Intel co-founder Gordon Moore surmised that transistor capacities evolved at a predictable and consistent rate. Specifically, all things considered, the number of transistors on a semiconductor chip would double about every two years. For a while, the so-called Moore’s Law held rather well—until, that is, things hit a literal wall.
“Moore’s Law was never meant to last forever,” according to a blog post by the Massachusetts Institute of Technology’s (MIT) Computer Science and Artificial Intelligence Lab. “Transistors can only get so small and, eventually, the more permanent laws of physics get in the way.” That is, as companies try to cram more transistors into smaller chips, new advances in transistor technology take longer than two years, so Moore’s Law has been over since at least 2016, Charles Leiserson, a computer scientist at MIT, said in the blog. Accordingly, the issue now is to consider how improvements in chip performance fit into a longer-term picture, Willy Shih, an economist at Harvard Business School, said in an explainer.
Reaching atomic levels In that sense, IBM’s latest chip represents an inventive approach for bypassing the limits of physical scaling. Specifically, two wafers with nanosheet-style transistors are glued together like a sandwich to vertically stack two layers of transistors, and related technical assessments suggested that the wafer stacking was flexible and scalable enough to support real computation, Huiming Bu, vice president of IBM’s silicon technology research team, said in a press briefing on the chip. Researcher holding IBM’s sub-1 nm node wafer. Credit: IBM That said, this chip isn’t quite ready for manufacturing just yet. The company’s goal is to enter production in the next five years, but there’s still work to be done. For instance, Bu pointed out that the team was still working on pathways to prevent thermal noise or integration into existing systems in the high-performance computing community. “From my perspective, I hope to see it be as successful as the 2-nanometer [chip] and become the industry platform,” Gambetta said during the briefing. “And as we see with AI and classical computing in general, we are only seeing more and more consumption.” #IBM #Crosses #Computings #Biggest #Barriers #Worlds #Sub1 #Nanometer #ChipIBM,Semiconductors,transistors](https://gizmodo.com/app/uploads/2026/06/nanostacking-ibm-sub-nm-chip-1280x720.jpg "IBM Crosses One of Computing’s Biggest Barriers With World’s First Sub-1 Nanometer Chip
In a major breakthrough, IBM revealed the world’s first semiconductor chip technology built on a sub-1 nanometer chipmaking process. For comparison, the process uses transistor features smaller than the width of a DNA strand, which measures about 2.5 nanometers across. The chip itself is about the size of a fingernail but holds almost 100 billion transistors, and the company expects it could enter markets as early as the next five years. In a statement released today, IBM said the new chip features nearly twice the density of its 2-nanometer chip, released in 2021. According to an accompanying technical report, the chip also demonstrated up to 70% greater energy efficiency than its predecessor. In designing the chip, researchers developed an “entirely new transistor architecture” called nanostack, which “vertically stacks and staggers transistors” to enable IBM’s 0.7-nanometer chip technology, IBM explained. A section of the chip seen with a transmission electron microscope. Credit: IBM “With our new nanostack architecture, we’re not just making smaller transistors,” Jay Gambetta, director of IBM Research, said in the statement. “We’re reinventing how chips are built to deliver dramatically more power and energy efficiency.”
Smaller and smaller Semiconductor chips enable things like computers, home appliances, communications, and transportation devices. In 1965, Intel co-founder Gordon Moore surmised that transistor capacities evolved at a predictable and consistent rate. Specifically, all things considered, the number of transistors on a semiconductor chip would double about every two years. For a while, the so-called Moore’s Law held rather well—until, that is, things hit a literal wall.
“Moore’s Law was never meant to last forever,” according to a blog post by the Massachusetts Institute of Technology’s (MIT) Computer Science and Artificial Intelligence Lab. “Transistors can only get so small and, eventually, the more permanent laws of physics get in the way.” That is, as companies try to cram more transistors into smaller chips, new advances in transistor technology take longer than two years, so Moore’s Law has been over since at least 2016, Charles Leiserson, a computer scientist at MIT, said in the blog. Accordingly, the issue now is to consider how improvements in chip performance fit into a longer-term picture, Willy Shih, an economist at Harvard Business School, said in an explainer.
Reaching atomic levels In that sense, IBM’s latest chip represents an inventive approach for bypassing the limits of physical scaling. Specifically, two wafers with nanosheet-style transistors are glued together like a sandwich to vertically stack two layers of transistors, and related technical assessments suggested that the wafer stacking was flexible and scalable enough to support real computation, Huiming Bu, vice president of IBM’s silicon technology research team, said in a press briefing on the chip. Researcher holding IBM’s sub-1 nm node wafer. Credit: IBM That said, this chip isn’t quite ready for manufacturing just yet. The company’s goal is to enter production in the next five years, but there’s still work to be done. For instance, Bu pointed out that the team was still working on pathways to prevent thermal noise or integration into existing systems in the high-performance computing community. “From my perspective, I hope to see it be as successful as the 2-nanometer [chip] and become the industry platform,” Gambetta said during the briefing. “And as we see with AI and classical computing in general, we are only seeing more and more consumption.” #IBM #Crosses #Computings #Biggest #Barriers #Worlds #Sub1 #Nanometer #ChipIBM,Semiconductors,transistors")
![IBM Crosses One of Computing’s Biggest Barriers With World’s First Sub-1 Nanometer Chip
In a major breakthrough, IBM revealed the world’s first semiconductor chip technology built on a sub-1 nanometer chipmaking process. For comparison, the process uses transistor features smaller than the width of a DNA strand, which measures about 2.5 nanometers across. The chip itself is about the size of a fingernail but holds almost 100 billion transistors, and the company expects it could enter markets as early as the next five years. In a statement released today, IBM said the new chip features nearly twice the density of its 2-nanometer chip, released in 2021. According to an accompanying technical report, the chip also demonstrated up to 70% greater energy efficiency than its predecessor. In designing the chip, researchers developed an “entirely new transistor architecture” called nanostack, which “vertically stacks and staggers transistors” to enable IBM’s 0.7-nanometer chip technology, IBM explained. A section of the chip seen with a transmission electron microscope. Credit: IBM “With our new nanostack architecture, we’re not just making smaller transistors,” Jay Gambetta, director of IBM Research, said in the statement. “We’re reinventing how chips are built to deliver dramatically more power and energy efficiency.”
Smaller and smaller Semiconductor chips enable things like computers, home appliances, communications, and transportation devices. In 1965, Intel co-founder Gordon Moore surmised that transistor capacities evolved at a predictable and consistent rate. Specifically, all things considered, the number of transistors on a semiconductor chip would double about every two years. For a while, the so-called Moore’s Law held rather well—until, that is, things hit a literal wall.
“Moore’s Law was never meant to last forever,” according to a blog post by the Massachusetts Institute of Technology’s (MIT) Computer Science and Artificial Intelligence Lab. “Transistors can only get so small and, eventually, the more permanent laws of physics get in the way.” That is, as companies try to cram more transistors into smaller chips, new advances in transistor technology take longer than two years, so Moore’s Law has been over since at least 2016, Charles Leiserson, a computer scientist at MIT, said in the blog. Accordingly, the issue now is to consider how improvements in chip performance fit into a longer-term picture, Willy Shih, an economist at Harvard Business School, said in an explainer.
Reaching atomic levels In that sense, IBM’s latest chip represents an inventive approach for bypassing the limits of physical scaling. Specifically, two wafers with nanosheet-style transistors are glued together like a sandwich to vertically stack two layers of transistors, and related technical assessments suggested that the wafer stacking was flexible and scalable enough to support real computation, Huiming Bu, vice president of IBM’s silicon technology research team, said in a press briefing on the chip. Researcher holding IBM’s sub-1 nm node wafer. Credit: IBM That said, this chip isn’t quite ready for manufacturing just yet. The company’s goal is to enter production in the next five years, but there’s still work to be done. For instance, Bu pointed out that the team was still working on pathways to prevent thermal noise or integration into existing systems in the high-performance computing community. “From my perspective, I hope to see it be as successful as the 2-nanometer [chip] and become the industry platform,” Gambetta said during the briefing. “And as we see with AI and classical computing in general, we are only seeing more and more consumption.” #IBM #Crosses #Computings #Biggest #Barriers #Worlds #Sub1 #Nanometer #ChipIBM,Semiconductors,transistors](https://i1.wp.com/gizmodo.com/app/uploads/2026/06/nanostacking-ibm-sub-nm-chip-1280x720.jpg?ssl=1 "IBM Crosses One of Computing’s Biggest Barriers With World’s First Sub-1 Nanometer Chip
In a major breakthrough, IBM revealed the world’s first semiconductor chip technology built on a sub-1 nanometer chipmaking process. For comparison, the process uses transistor features smaller than the width of a DNA strand, which measures about 2.5 nanometers across. The chip itself is about the size of a fingernail but holds almost 100 billion transistors, and the company expects it could enter markets as early as the next five years. In a statement released today, IBM said the new chip features nearly twice the density of its 2-nanometer chip, released in 2021. According to an accompanying technical report, the chip also demonstrated up to 70% greater energy efficiency than its predecessor. In designing the chip, researchers developed an “entirely new transistor architecture” called nanostack, which “vertically stacks and staggers transistors” to enable IBM’s 0.7-nanometer chip technology, IBM explained. A section of the chip seen with a transmission electron microscope. Credit: IBM “With our new nanostack architecture, we’re not just making smaller transistors,” Jay Gambetta, director of IBM Research, said in the statement. “We’re reinventing how chips are built to deliver dramatically more power and energy efficiency.”
Smaller and smaller Semiconductor chips enable things like computers, home appliances, communications, and transportation devices. In 1965, Intel co-founder Gordon Moore surmised that transistor capacities evolved at a predictable and consistent rate. Specifically, all things considered, the number of transistors on a semiconductor chip would double about every two years. For a while, the so-called Moore’s Law held rather well—until, that is, things hit a literal wall.
“Moore’s Law was never meant to last forever,” according to a blog post by the Massachusetts Institute of Technology’s (MIT) Computer Science and Artificial Intelligence Lab. “Transistors can only get so small and, eventually, the more permanent laws of physics get in the way.” That is, as companies try to cram more transistors into smaller chips, new advances in transistor technology take longer than two years, so Moore’s Law has been over since at least 2016, Charles Leiserson, a computer scientist at MIT, said in the blog. Accordingly, the issue now is to consider how improvements in chip performance fit into a longer-term picture, Willy Shih, an economist at Harvard Business School, said in an explainer.
Reaching atomic levels In that sense, IBM’s latest chip represents an inventive approach for bypassing the limits of physical scaling. Specifically, two wafers with nanosheet-style transistors are glued together like a sandwich to vertically stack two layers of transistors, and related technical assessments suggested that the wafer stacking was flexible and scalable enough to support real computation, Huiming Bu, vice president of IBM’s silicon technology research team, said in a press briefing on the chip. Researcher holding IBM’s sub-1 nm node wafer. Credit: IBM That said, this chip isn’t quite ready for manufacturing just yet. The company’s goal is to enter production in the next five years, but there’s still work to be done. For instance, Bu pointed out that the team was still working on pathways to prevent thermal noise or integration into existing systems in the high-performance computing community. “From my perspective, I hope to see it be as successful as the 2-nanometer [chip] and become the industry platform,” Gambetta said during the briefing. “And as we see with AI and classical computing in general, we are only seeing more and more consumption.” #IBM #Crosses #Computings #Biggest #Barriers #Worlds #Sub1 #Nanometer #ChipIBM,Semiconductors,transistors")
-SOURCE-Scott-Gilbertson.jpg)
-SOURCE-Scott-Gilbertson.jpg)
-SOURCE-Scott-Gilbertson.jpg)
Post Comment