#gpus — Public Fediverse posts

@dannotdaniel @GhostOnTheHalfShell Great video if you want to understand the current #AI bubble and why it is inevitably going to burst, and soon. It is very complicated to understand, but here are my takeaways. The answers are only partially that AI actual capability being oversold. The main reasons are:

1. Data centers are very difficult to build, even if you have the power and water available (which they don’t). There are lots of things that can go wrong, and are, and will. The estimate is only 10% of those under contract will EVER become operational.

2. The #hyperscaler companies (#Oracle) who contract to build data centers , and their suppliers, have taken on enormous debt that is putting these S&P 500 companies at risk for junk bond status.

3. The chip providers (#NVidia for example) have oversold and over supplied. Those #GPUs you can’t get either have not been built or are sitting in boxes in warehouses.

4. Demand for AI services is not based on a reality where end-user companies are making boatloads of money. It is based on fear of missing out.

@dannotdaniel @GhostOnTheHalfShell Great video if you want to understand the current #AI bubble and why it is inevitably going to burst, and soon. It is very complicated to understand, but here are my takeaways. The answers are only partially that AI actual capability being oversold. The main reasons are:

1. Data centers are very difficult to build, even if you have the power and water available (which they don’t). There are lots of things that can go wrong, and are, and will. The estimate is only 10% of those under contract will EVER become operational.

2. The #hyperscaler companies (#Oracle) who contract to build data centers , and their suppliers, have taken on enormous debt that is putting these S&P 500 companies at risk for junk bond status.

3. The chip providers (#NVidia for example) have oversold and over supplied. Those #GPUs you can’t get either have not been built or are sitting in boxes in warehouses.

4. Demand for AI services is not based on a reality where end-user companies are making boatloads of money. It is based on fear of missing out.

@dannotdaniel @GhostOnTheHalfShell Great video if you want to understand the current #AI bubble and why it is inevitably going to burst, and soon. It is very complicated to understand, but here are my takeaways. The answers are only partially that AI actual capability being oversold. The main reasons are:

1. Data centers are very difficult to build, even if you have the power and water available (which they don’t). There are lots of things that can go wrong, and are, and will. The estimate is only 10% of those under contract will EVER become operational.

2. The #hyperscaler companies (#Oracle) who contract to build data centers , and their suppliers, have taken on enormous debt that is putting these S&P 500 companies at risk for junk bond status.

3. The chip providers (#NVidia for example) have oversold and over supplied. Those #GPUs you can’t get either have not been built or are sitting in boxes in warehouses.

4. Demand for AI services is not based on a reality where end-user companies are making boatloads of money. It is based on fear of missing out.

@dannotdaniel @GhostOnTheHalfShell Great video if you want to understand the current #AI bubble and why it is inevitably going to burst, and soon. It is very complicated to understand, but here are my takeaways. The answers are only partially that AI actual capability being oversold. The main reasons are:

1. Data centers are very difficult to build, even if you have the power and water available (which they don’t). There are lots of things that can go wrong, and are, and will. The estimate is only 10% of those under contract will EVER become operational.

2. The #hyperscaler companies (#Oracle) who contract to build data centers , and their suppliers, have taken on enormous debt that is putting these S&P 500 companies at risk for junk bond status.

3. The chip providers (#NVidia for example) have oversold and over supplied. Those #GPUs you can’t get either have not been built or are sitting in boxes in warehouses.

4. Demand for AI services is not based on a reality where end-user companies are making boatloads of money. It is based on fear of missing out.