Join our daily and weekly newsletters for the latest updates and exclusive content on industry-leading AI coverage. Learn More UK-based chip designer Arm offers the architecture for systems-on-a-chip (SoCs) that are used by some of the world’s largest tech brands, from Nvidia to Amazon to Google parent company Alphabet and beyond, all without ever manufacturing any hardware of its own — though that’s reportedly due to change this year. And you’d think with a record setting last quarter of $1.24 billion in total revenue, it might want to just keep things steady and keep raking in the cash. But Arm sees how fast AI has taken off in the enterprise, and with some of its customers delivering record revenue of their own by offering AI graphics processing units that incorporate Arm’s tech, Arm wants a piece of the action. Today, the company announced a new product naming strategy that underscores its shift from a supplier of component IP to a platform-first company. “It’s about showing customers that we have much more to offer than just hardware and chip designs. specifically — we have a whole ecosystem that can help them scale AI and do so at lower cost with greater efficiency,” said Arm’s chief marketing officer Ami Badani, in an exclusive interview with VentureBeat over Zoom yesterday. Indeed, as Arm CEO Rene Haas told the tech news outlet Next Platform back in February, Arm’s history of creating lower-power chips than the competition (cough cough, Intel) has set it up extremely well to serve as the basis for power-hungry AI training and inference jobs. According to his comments in that article, today’s data center consume approximately 460 terawatt hours of electricity per year, but that is expected to triple by the end of this decade, and could jump from being 4 percent of all of the world’s energy usage to 25 percent — unless more Arm power-saving chip designs and their accompanying optimized software and firmware are used in the infrastructure for these centers. From IP to platform: a significant shift As AI workloads scale in complexity and power requirements, Arm is reorganizing its offerings around complete compute platforms. These platforms allow for faster integration, more efficient scaling, and lower complexity for partners building AI-capable chips. To reflect this shift, Arm is retiring its prior naming conventions and introducing new product families that are organized by market: Neoverse for infrastructure Niva for PCs Lumex for mobile Zena for automotive Orbis for IoT and edge AI The Mali brand will continue to represent GPU offerings, integrated as components within these new platforms. Alongside the renaming, Arm is overhauling its product numbering system. IP identifiers will now correspond to platform generations and performance tiers labeled Ultra, Premium, Pro, Nano, and Pico. This structure is aimed at making the roadmap more transparent to customers and developers. Emboldened by strong results The rebranding follows Arm’s strong Q4 fiscal year 2025 (ended March 31), where the company crossed the $1 billion mark in quarterly revenue for the first time. Total revenue hit $1.24 billion, up 34% year-over-year, driven by both record licensing revenue ($634 million, up 53%) and royalty revenue ($607 million, up 18%). Notably, this royalty growth was driven by increasing deployment of the Armv9 architecture and adoption of Arm Compute Subsystems (CSS) across smartphones, cloud infrastructure, and edge AI. The mobile market was a standout: while global smartphone shipments grew less than 2%, Arm’s smartphone royalty revenue rose roughly 30%. The company also entered its first automotive CSS agreement with a leading global EV manufacturer, furthering its penetration into the high-growth automotive market. While Arm hasn’t disclosed the EV manufacturer’s precise name yet, Badani told VentureBeat that it sees automotive as a major growth area in addition to AI model providers and cloud hyperscalers such as Google and Amazon. “We’re looking at automotive as a major growth area and we believe that AI and other advances like self-driving are going to be standard, which our designs are perfect for,” the CMO told VentureBeat. Meanwhile, cloud providers like AWS, Google Cloud, and Microsoft Azure continued expanding their use of Arm-based silicon to run AI workloads, affirming Arm’s growing influence in data center compute. Growing a new platform ecosystem with software and vertically integrated products Arm is complementing its hardware platforms with expanded software tools and ecosystem support. Its extension for GitHub Copilot, now free for all developers, lets users optimize code using Arm’s architecture. More than 22 million developers now build on Arm, and its Kleidi AI software layer has surpassed 8 billion cumulative installs across devices. Arm’s leadership sees the rebrand as a natural step in its long-term strategy. By providing vertically integrated platforms with performance and naming clarity, the company aims to meet increasing demand for energy-efficient AI compute from device to data center. As Haas wrote in Arm’s blog post, Arm’s compute platforms are foundational to a future where AI is everywhere—and Arm is poised to deliver that foundation at scale. What it means for AI and data decision makers This strategic repositioning is likely to reshape how technical decision makers across AI, data, and security roles approach their day-to-day work and future planning. For those managing large language model lifecycles, the clearer platform structure offers a more streamlined path for selecting compute architectures optimized for AI workloads. As model deployment timelines tighten and the bar for efficiency rises, having predefined compute systems like Neoverse or Lumex could reduce the overhead required to evaluate raw IP blocks and allow faster execution in iterative development cycles. For engineers orchestrating AI pipelines across environments, the modularity and performance tiering within Arm’s new architecture could help simplify pipeline standardization. It introduces a practical way to align compute capabilities with varying workload requirements—whether that’s running inference at the edge or managing resource-intensive training jobs in the cloud. These engineers, often juggling system uptime and cost-performance tradeoffs, may find more clarity in mapping their orchestration logic to predefined Arm platform tiers. Data infrastructure leaders tasked with maintaining high-throughput pipelines and ensuring data integrity may also benefit. The naming update and system-level integration signal a deeper commitment from Arm to support scalable designs that work well with AI-enabled pipelines. The compute subsystems may also accelerate time-to-market for custom silicon that supports next-gen data platforms—important for teams that operate under budget constraints and limited engineering bandwidth. Security leaders, meanwhile, will likely see implications in how embedded security features and system-level compatibility evolve within these platforms. With Arm aiming to offer consistent architecture across edge and cloud, security teams can more easily plan for and enforce end-to-end protections, especially when integrating AI workloads that demand both performance and strict access controls. The broader effect of this branding shift is a signal to enterprise architects and engineers: Arm is no longer just a component provider—it’s offering full-stack foundations for how AI systems are built and scaled. Daily insights on business use cases with VB Daily If you want to impress your boss, VB Daily has you covered. We give you the inside scoop on what companies are doing with generative AI, from regulatory shifts to practical deployments, so you can share insights for maximum ROI. Read our Privacy Policy Thanks for subscribing. Check out more VB newsletters here. An error occured.

The saying goes that 80% of data collected, stored and maintained by governments can be associated with geographical locations. Although never empirically proven, it illustrates the importance of location within data. Ever growing data volumes put constraints on systems that handle geospatial data. Common Big Data compute engines, originally designed to scale for textual data, need adaptation to work efficiently with geospatial data — think of geographical indexes, partitioning, and operators. Here, I present and illustrate how to utilize the Microsoft Fabric Spark compute engine, with the natively integrated ESRI GeoAnalytics engine# for geospatial big data processing and analytics. The optional GeoAnalytics capabilities within Fabric enable the processing and analytics of vector-type geospatial data, where vector-type geospatial data refers to points, lines, polygons. These capabilities include more than 150 spatial functions to create geometries, test, and select spatial relationships. As it extends Spark, the GeoAnalytics functions can be called when using Python, SQL, or Scala. These spatial operations apply automatically spatial indexing, making the Spark compute engine also efficient for this data. It can handle 10 extra common spatial data formats to load and save data spatial data, on top of the Spark natively supported data source formats. This blog post focuses on the scalable geospatial compute engines as has been introduced in my post about geospatial in the age of AI. Demonstration explained Here, I demonstrate some of these spatial capabilities by showing the data manipulation and analytics steps on a large dataset. By using several tiles covering point cloud data (a bunch of x, y, z values), an enormous dataset starts to form, while it still covers a relatively small area. The open Dutch AHN dataset, which is a national digital elevation and surface model, is currently in its fifth update cycle, and spans a period of nearly 30 years. Here, the data from the second, third, and forth acquisition is used, as these hold full national coverage (the fifth just not yet), while the first version did not include a point cloud release (only the derivative gridded version). Another Dutch open dataset, namely building data, the BAG, is used to illustrate spatial selection. The building dataset contains the footprint of the buildings as polygons. Currently, this dataset holds more than 11 million buildings. To test the spatial functions, I use only 4 AHN tiles per AHN version. Thus in this case, 12 tiles, each of 5 x 6.25 km. Totalling to more than 3.5 billion points within an area of 125 square kilometers. The chosen area covers the municipality of Loppersum, an area prone to land subsidence due to gas extraction. The steps to take include the selection of buildings within the area of Loppersum, selecting the x,y,z-points from the roofs of the buildings. Then, we bring the 3 datasets into one dataframe and do an extra analysis with it. A spatial regression to predict the expected height of a building based on its height history as well as the history of the buildings in its direct surroundings. Not necessarily the best analysis to perform on this data to come to actual predictions* but it suits merely the purpose of demonstrating the spatial processing capabilities of Fabric’s ESRI GeoAnalytics. All the below code snippets are also available as notebooks on github. Step 1: Read data Spatial data can come in many different data formats; we conform to the geoparquet data format for further processing. The BAG building data, both the footprints as well as the accompanied municipality boundaries, come in geoparquet format already. The point cloud AHN data, version 2, 3 and 4, however, comes as LAZ file formats — a compressed industry standard format for point clouds. I have not found a Spark library to read LAZ (please leave a message in case there is one), and created a txt file, separately, with the LAStools+ first. # ESRI - FABRIC reference: https://developers.arcgis.com/geoanalytics-fabric/ # Import the required modules import geoanalytics_fabric from geoanalytics_fabric.sql import functions as ST from geoanalytics_fabric import extensions # Read ahn file from OneLake # AHN lidar data source: https://viewer.ahn.nl/ ahn_csv_path = "Files/AHN lidar/AHN4_csv" lidar_df = spark.read.options(delimiter=" ").csv(ahn_csv_path) lidar_df = lidar_df.selectExpr("_c0 as X", "_c1 as Y", "_c2 Z") lidar_df.printSchema() lidar_df.show(5) lidar_df.count() The above code snippet& provides the below results: Now, with the spatial functions make_point and srid the x,y,z columns are transformed to a point geometry and set it to the specific Dutch coordinate system (SRID = 28992), see the below code snippet&: # Create point geometry from x,y,z columns and set the spatial refrence system lidar_df = lidar_df.select(ST.make_point(x="X", y="Y", z="Z").alias("rd_point")) lidar_df = lidar_df.withColumn("srid", ST.srid("rd_point")) lidar_df = lidar_df.select(ST.srid("rd_point", 28992).alias("rd_point"))\ .withColumn("srid", ST.srid("rd_point")) lidar_df.printSchema() lidar_df.show(5) Building and municipality data can be read with the extended spark.read function for geoparquet, see the code snippet&: # Read building polygon data path_building = "Files/BAG NL/BAG_pand_202504.parquet" df_buildings = spark.read.format("geoparquet").load(path_building) # Read woonplaats data (=municipality) path_woonplaats = "Files/BAG NL/BAG_woonplaats_202504.parquet" df_woonplaats = spark.read.format("geoparquet").load(path_woonplaats) # Filter the DataFrame where the "woonplaats" column contains the string "Loppersum" df_loppersum = df_woonplaats.filter(col("woonplaats").contains("Loppersum")) Step 2: Make selections In the accompanying notebooks, I read and write to geoparquet. To make sure the right data is read correctly as dataframes, see the following code snippet: # Read building polygon data path_building = "Files/BAG NL/BAG_pand_202504.parquet" df_buildings = spark.read.format("geoparquet").load(path_building) # Read woonplaats data (=municipality) path_woonplaats = "Files/BAG NL/BAG_woonplaats_202504.parquet" df_woonplaats = spark.read.format("geoparquet").load(path_woonplaats) # Filter the DataFrame where the "woonplaats" column contains the string "Loppersum" df_loppersum = df_woonplaats.filter(col("woonplaats").contains("Loppersum")) With all data in dataframes it becomes a simple step to do spatial selections. The following code snippet& shows how to select the buildings within the boundaries of the Loppersum municipality, and separately makes a selection of buildings that existed throughout the period (point cloud AHN-2 data was acquired in 2009 in this region). This resulted in 1196 buildings, out of the 2492 buildings currently. # Clip the BAG buildings to the gemeente Loppersum boundary df_buildings_roi = Clip().run(input_dataframe=df_buildings, clip_dataframe=df_loppersum) # select only buildings older then AHN data (AHN2 (Groningen) = 2009) # and with a status in use (Pand in gebruik) df_buildings_roi_select = df_buildings_roi.where((df_buildings_roi.bouwjaar<2009) & (df_buildings_roi.status=='Pand in gebruik')) The three AHN versions used (2,3 and 4), further named as T1, T2 and T3 respectively, are then clipped based on the selected building data. The AggregatePoints function can be utilized to calculate, in this case from the height (z-values) some statistics, like the mean per roof, the standard deviation and the number of z-values it is based upon; see the code snippet: # Select and aggregrate lidar points from buildings within ROI df_ahn2_result = AggregatePoints() \ .setPolygons(df_buildings_roi_select) \ .addSummaryField(summary_field="T1_z", statistic="Mean", alias="T1_z_mean") \ .addSummaryField(summary_field="T1_z", statistic="stddev", alias="T1_z_stddev") \ .run(df_ahn2) df_ahn3_result = AggregatePoints() \ .setPolygons(df_buildings_roi_select) \ .addSummaryField(summary_field="T2_z", statistic="Mean", alias="T2_z_mean") \ .addSummaryField(summary_field="T2_z", statistic="stddev", alias="T2_z_stddev") \ .run(df_ahn3) df_ahn4_result = AggregatePoints() \ .setPolygons(df_buildings_roi_select) \ .addSummaryField(summary_field="T3_z", statistic="Mean", alias="T3_z_mean") \ .addSummaryField(summary_field="T3_z", statistic="stddev", alias="T3_z_stddev") \ .run(df_ahn4) Step 3: Aggregate and Regress As the GeoAnalytics function Geographically Weighted Regression (GWR) can only work on point data, from the building polygons their centroid is extracted with the centroid function. The 3 dataframes are joined to one, see also the notebook, and it is ready to perform the GWR function. In this instance, it predicts the height for T3 (AHN4) based on local regression functions. # Import the required modules from geoanalytics_fabric.tools import GWR # Run the GWR tool to predict AHN4 (T3) height values for buildings at Loppersum resultGWR = GWR() \ .setExplanatoryVariables("T1_z_mean", "T2_z_mean") \ .setDependentVariable(dependent_variable="T3_z_mean") \ .setLocalWeightingScheme(local_weighting_scheme="Bisquare") \ .setNumNeighbors(number_of_neighbors=10) \ .runIncludeDiagnostics(dataframe=df_buildingsT123_points) The model diagnostics can be consulted for the predicted z value, in this case, the following results were generated. Note, again, that these results cannot be used for real world applications as the data and methodology might not best fit the purpose of subsidence modelling — it merely shows here Fabric GeoAnalytics functionality. R20.994AdjR20.981AICc1509Sigma20.046EDoF378 Step 4: Visualize results With the spatial function plot, results can be visualized as maps within the notebook — to be used only with the Python API in Spark. First, a visualization of all buildings within the municipality of Loppersum. # visualize Loppersum buildings df_buildings.st.plot(basemap="light", geometry="geometry", edgecolor="black", alpha=0.5) Here is a visualization of the height difference between T3 (AHN4) and T3 predicted (T3 predicted minus T3). # Vizualize difference of predicted height and actual measured height Loppersum area and buildings axes = df_loppersum.st.plot(basemap="light", edgecolor="black", figsize=(7, 7), alpha=0) axes.set(xlim=(244800, 246500), ylim=(594000, 595500)) df_buildings.st.plot(ax=axes, basemap="light", alpha=0.5, edgecolor="black") #, color='xkcd:sea blue' df_with_difference.st.plot(ax=axes, basemap="light", cmap_values="subsidence_mm_per_yr", cmap="coolwarm_r", vmin=-10, vmax=10, geometry="geometry") Summary This blog post discusses the significance of geographical data. It highlights the challenges posed by increasing data volumes on Geospatial data systems and suggests that traditional big data engines must adapt to handle geospatial data efficiently. Here, an example is presented on how to use the Microsoft Fabric Spark compute engine and its integration with the ESRI GeoAnalytics engine for effective geospatial big data processing and analytics. Opinions here are mine. Footnotes # in preview * for modelling the land subsidence with much higher accuracy and temporal frequency other approaches and data can be utilized, such as with satellite InSAR methodology (see also Bodemdalingskaart) + Lastools is used here separately, it would be fun to test the usage of Fabric User data functions (preview), or to utilize an Azure Function for this purpose. & code snippets here are set up for readability, not necessarily for efficiency. Multiple data processing steps could be chained. References GitHub repo with notebooks: delange/Fabric_GeoAnalytics Microsoft Fabric: Microsoft Fabric documentation – Microsoft Fabric | Microsoft Learn ESRI GeoAnalytics for Fabric: Overview | ArcGIS GeoAnalytics for Microsoft Fabric | ArcGIS Developers AHN: Home | AHN BAG: Over BAG – Basisregistratie Adressen en Gebouwen – Kadaster.nl zakelijk Lastools: LAStools: converting, filtering, viewing, processing, and compressing LIDAR data in LAS and LAZ format Surface and Object Motion Map: Bodemdalingskaart – The post The Geospatial Capabilities of Microsoft Fabric and ESRI GeoAnalytics, Demonstrated appeared first on Towards Data Science.

Microsoft quietly announced earlier this week that it plans to shut down a long-standing tool supplying search engine startups and other software developers with a raw feed of Bing search results. From a report: The Bing Search APIs, or application programming interfaces, were once vital to many niche Google alternatives, but fell out of favor more recently as Microsoft hiked fees for the service and restricted its use. The shutoff, which is scheduled to begin on August 11, still came as a surprise to several developers who spoke with WIRED. Customers learned of it on Monday via an email from Microsoft and a post on its website. They were directed to consider using "Grounding with Bing Search as part of Azure AI Agents," a Microsoft service that allows chatbots like ChatGPT to augment AI-generated responses with "real-time public web data." Some developers view the AI-centric alternative as an insufficient replacement. Larger customers like DuckDuckGo told Wired they won't be affected. Read more of this story at Slashdot.

With the amount of Data growing exponentially in the last few years, one of the biggest challenges has become finding the most optimal way to store various data flavors. Unlike in the (not so far) past, when relational databases were considered the only way to go, organizations now want to perform analysis over raw data – think of social media sentiment analysis, audio/video files, and so on – which usually couldn’t be stored in a traditional (relational) way, or storing them in a traditional way would require significant effort and time, which increase the overall time-for-analysis. Another challenge was to somehow stick with a traditional approach to have data stored in a structured way, but without the necessity to design complex and time-consuming ETL workloads to move this data into the enterprise data warehouse. Additionally, what if half of the data professionals in your organization are proficient with, let’s say, Python (data scientists, data engineers), and the other half (data engineers, data analysts) with SQL? Would you insist that “Pythonists” learn SQL? Or, vice-versa? Or, would you prefer a storage option that can play to the strengths of your entire data team? I have good news for you – something like this has already existed since 2013, and it’s called Apache Parquet! Parquet file format in a nutshell Before I show you the ins and outs of the Parquet file format, there are (at least) five main reasons why Parquet is considered a de facto standard for storing data nowadays: Data compression – by applying various encoding and compression algorithms, Parquet file provides reduced memory consumption Columnar storage – this is of paramount importance in analytic workloads, where fast data read operation is the key requirement. But, more on that later in the article… Language agnostic – as already mentioned previously, developers may use different programming languages to manipulate the data in the Parquet file Open-source format – meaning, you are not locked with a specific vendor Support for complex data types Row-store vs Column-store We’ve already mentioned that Parquet is a column-based storage format. However, to understand the benefits of using the Parquet file format, we first need to draw the line between the row-based and column-based ways of storing the data. In traditional, row-based storage, the data is stored as a sequence of rows. Something like this: Image by author Now, when we are talking about OLAP scenarios, some of the common questions that your users may ask are: How many balls did we sell? How many users from the USA bought a T-shirt? What is the total amount spent by customer Maria Adams? How many sales did we have on January 2nd? To be able to answer any of these questions, the engine must scan each and every row from the beginning to the very end! So, to answer the question: how many users from the USA bought T-shirt, the engine has to do something like this: Image by author Essentially, we just need the information from two columns: Product (T-Shirts) and Country (USA), but the engine will scan all five columns! This is not the most efficient solution – I think we can agree on that… Column store Let’s now examine how the column store works. As you may assume, the approach is 180 degrees different: Image by author In this case, each column is a separate entity – meaning, each column is physically separated from other columns! Going back to our previous business question: the engine can now scan only those columns that are needed by the query (Product and country), while skipping scanning the unnecessary columns. And, in most cases, this should improve the performance of the analytical queries. Ok, that’s nice, but the column store existed before Parquet and it still exists outside of Parquet as well. So, what is so special about the Parquet format? Parquet is a columnar format that stores the data in row groups Wait, what?! Wasn’t it complicated enough even before this? Don’t worry, it’s much easier than it sounds Let’s go back to our previous example and depict how Parquet will store this same chunk of data: Image by author Let’s stop for a moment and explain the illustration above, as this is exactly the structure of the Parquet file (some additional things were intentionally omitted, but we will come soon to explain that as well). Columns are still stored as separate units, but Parquet introduces additional structures, called Row group. Why is this additional structure super important? You’ll need to wait for an answer for a bit :). In OLAP scenarios, we are mainly concerned with two concepts: projection and predicate(s). Projection refers to a SELECT statement in SQL language – which columns are needed by the query. Back to our previous example, we need only the Product and Country columns, so the engine can skip scanning the remaining ones. Predicate(s) refer to the WHERE clause in SQL language – which rows satisfy criteria defined in the query. In our case, we are interested in T-Shirts only, so the engine can completely skip scanning Row group 2, where all the values in the Product column equal socks! Image by author Let’s quickly stop here, as I want you to realize the difference between various types of storage in terms of the work that needs to be performed by the engine: Row store – the engine needs to scan all 5 columns and all 6 rows Column store – the engine needs to scan 2 columns and all 6 rows Column store with row groups – the engine needs to scan 2 columns and 4 rows Obviously, this is an oversimplified example, with only 6 rows and 5 columns, where you will definitely not see any difference in performance between these three storage options. However, in real life, when you’re dealing with much larger amounts of data, the difference becomes more evident. Now, the fair question would be: how does Parquet “know” which row group to skip/scan? Parquet file contains metadata This means that every Parquet file contains “data about data” – information such as minimum and maximum values in a specific column within a certain row group. Furthermore, every Parquet file contains a footer, which keeps the information about the format version, schema information, column metadata, and so on. You can find more details about Parquet metadata types here. Important: In order to optimize the performance and eliminate unnecessary data structures (row groups and columns), the engine first needs to “get familiar” with the data, so it first reads the metadata. It’s not a slow operation, but it still requires a certain amount of time. Therefore, if you’re querying the data from multiple small Parquet files, query performance can degrade, because the engine will have to read metadata from each file. So, you should be better off merging multiple smaller files into one bigger file (but still not too big :)… I hear you, I hear you: Nikola, what is “small” and what is “big”? Unfortunately, there is no single “golden” number here, but for example, Microsoft Azure Synapse Analytics recommends that the individual Parquet file should be at least a few hundred MBs in size. What else is in there? Here is a simplified, high-level illustration of the Parquet file format: Image by author Can it be better than this? Yes, with data compression Ok, we’ve explained how skipping the scan of the unnecessary data structures (row groups and columns) may benefit your queries and increase the overall performance. But, it’s not only about that – remember when I told you at the very beginning that one of the main advantages of the Parquet format is the reduced memory footprint of the file? This is achieved by applying various compression algorithms. I’ve already written about various data compression types in Power BI (and the Tabular model in general) here, so maybe it’s a good idea to start by reading this article. There are two main encoding types that enable Parquet to compress the data and achieve astonishing savings in space: Dictionary encoding – Parquet creates a dictionary of the distinct values in the column, and afterward replaces “real” values with index values from the dictionary. Going back to our example, this process looks something like this: Image by author You might think: why this overhead, when product names are quite short, right? Ok, but now imagine that you store the detailed description of the product, such as: “Long arm T-Shirt with application on the neck”. And, now imagine that you have this product sold million times…Yeah, instead of having million times repeating value “Long arm…bla bla”, the Parquet will store only the Index value (integer instead of text). Run-Length-Encoding with Bit-Packing – when your data contains many repeating values, Run-Length-Encoding (RLE) algorithm may bring additional memory savings. Can it be better than THIS?! Yes, with the Delta Lake file format Ok, what the heck is now a Delta Lake format?! This is the article about Parquet, right? So, to put it in plain English: Delta Lake is nothing else but the Parquet format “on steroids”. When I say “steroids”, the main one is the versioning of Parquet files. It also stores a transaction log to enable tracking all changes applied to the Parquet file. This is also known as ACID-compliant transactions. Since it supports not only ACID transactions, but also supports time travel (rollbacks, audit trails, etc.) and DML (Data Manipulation Language) statements, such as INSERT, UPDATE and DELETE, you won’t be wrong if you think of the Delta Lake as a “data warehouse on the data lake” (who said: Lakehouse). Examining the pros and cons of the Lakehouse concept is out of the scope of this article, but if you’re curious to go deeper into this, I suggest you read this article from Databricks. Conclusion We evolve! Same as we, the data is also evolving. So, new flavors of data required new ways of storing it. The Parquet file format is one of the most efficient storage options in the current data landscape, since it provides multiple benefits – both in terms of memory consumption, by leveraging various compression algorithms, and fast query processing by enabling the engine to skip scanning unnecessary data. Thanks for reading! The post Parquet File Format – Everything You Need to Know! appeared first on Towards Data Science.

Deal  When you purchase through links on our site, we may earn an affiliate commission. Here’s how it works. Buy Windows 11 Pro for 2 devices with a copy of Office 2021 Pro for only $55 Neowin Deals · May 14, 2025 13:00 EDT Today's highlighted deal comes via our Apps + Software section of the Neowin Deals store, where you can save 86% on Windows 11 Pro (for 2 devices) + Microsoft Office Pro 2021. Upgrade your computing experience with Windows 11 Pro. This cutting-edge operating system boasts a sleek new design and advanced tools to help you work faster and smarter. From creative projects to gaming and beyond, Windows 11 delivers the power and flexibility you need to achieve your goals. With a focus on productivity, the new features are easy to learn and use, enhancing your workflow and efficiency. Whether you're a student, professional, gamer, or creative, Windows 11 Home has everything you need to take your productivity to the next level. New interface. easier on the eyes & easier to use Biometrics login*.Encrypted authentication & advanced antivirus defenses DirectX 12 Ultimate. Play the latest games with graphics that rival reality. DirectX 12 Ultimate comes ready to maximize your hardware* Screen space. Snap layouts, desktops & seamless redocking Widgets. Stay up-to-date with the content you love & the new you care about Microsoft Teams. Stay in touch with friends and family with Microsoft Teams, which can be seamlessly integrated into your taskbar** Wake & lock. Automatically wake up when you approach and lock when you leave Smart App Control. Provides a layer of security by only permitting apps with good reputations to be installed Windows Studio Effects. Designed with Background Blur, Eye Contact, Voice Focus, & Automatic Framing Touchscreen. For a true mouse-less or keyboard-less experience TPM 2.0. Helps prevent unwanted tampering Windows 11 Pro also includes a number of productivity-focused features, such as the ability to snap multiple windows together and create custom layouts, improved voice typing, and a new, more powerful search experience. Personal and professional users will enjoy a modern and secure computing experience, with improved performance and productivity features to help users get more done. Only on Windows 11 Pro If you require enterprise-oriented features for your daily professional tasks, then Windows 11 Pro is a better option. Set up with a local account (only when set up for work or school) Join Active Directory/Azure AD Hyper-V Windows Sandbox Microsoft Remote Desktop BitLocker device encryption Windows Information Protection Mobile device management (MDM) Group Policy Enterprise State Roaming with Azure Assigned Access Dynamic Provisioning Windows Update for Business Kiosk mode Maximum RAM: 2TB Maximum no. of CPUs: 2 Maximum no. of CPU cores: 128 Good to know: Length of access: lifetime Redemption deadline: redeem your code within 30 days of purchase Access options: desktop Max number of device(s): 2 (Use one activation key for up to 2 devices) Version: Windows 11 Pro Updates included Click here to verify Microsoft partnership For Windows 10 or Newer! Get All Essential Microsoft Apps for Your PC with This One-Time PurchaseThis is intended for families and small businesses who want classic Office apps and email. It includes Word, Excel, PowerPoint, Outlook, Teams, and OneNote. A one-time purchase installed on 1 Windows PC for use at home or work. Lifetime license for MS Word, Excel, PowerPoint, Outlook, Teams, & OneNote One-time purchase installed on 1 Windows PC for use at home or work Instant Delivery & Download – access your software license keys and download links instantly Free customer service – only the best support! Microsoft Office Professional 2021 (for Windows) includes: Microsoft Office Word Microsoft Office Excel Microsoft Office PowerPoint Microsoft Office Outlook Microsoft Office Teams Microsoft Office OneNote Microsoft Office Publisher Microsoft Office Access Good to know: ONE-TIME PURCHASE INSTALLED ON 1 DEVICE Redemption deadline: redeem your code within 30 days of purchase Access options: desktop Full versions No subscriptions – no monthly/annual fees Version: 2021 Updates included Here's the deal: This Microsoft Office Pro 2021 + Windows 11 Pro normally costs $438, but this deal can be yours from just $54.97, that's a saving of $383. For full terms, specifications, and license info please click the link below. Get this deal for just $54.97 (86% off), or learn more See all discounted Neowin Deals on offer. This is a time-limited deal.Although priced in U.S. dollars, this deal is available for digital purchase worldwide. We post these because we earn commission on each sale so as not to rely solely on advertising, which many of our readers block. It all helps toward paying staff reporters, servers and hosting costs. Other ways to support Neowin Whitelist Neowin by not blocking our ads Create a free member account to see fewer ads Subscribe to Neowin - for $14 a year, or $28 a year for an ad-free experience Disclosure: Neowin benefits from revenue of each sale made through our branded deals site powered by StackCommerce. Tags Report a problem with article Follow @NeowinFeed

Amazon has restocked Pokémon TCG: Prismatic Evolutions Surprise Boxes today... for $59.99. But is it worth it? Well, in the TCG community right now, you've got MSRP vigilants ($22.99), and those willing to pay at "Market Price" ($59.99). This Surprise Box is sold and shipped by Amazon, so it's what you'd consider market price; it's just way above MSRP. Whatever side of the fence you sit on, the debate is fascinating.Pokémon TCG: Prismatic Evolutions Surprise BoxIncludes 1 of 9 randomly selected foil promo cards with a special logo. Warning: MSRP: $22.99.$59.99 at AmazonLet's dig into whether it's worth paying $59.99 for this set. For starters, the box itself is nice, and it's cool to get stamped versions of the standard Eeveelution ex cards: Eevee ex, Vaporeon ex, Jolteon ex, Flareon ex, Espeon ex, Umbreon ex, Leafeon ex, Glaceon ex, or Sylveon ex.But you also only get one per box at random, and I'm not going to give myself an anxiety attack trying to work out the odds on hunting all of these down, but that's a big bag of bones going off Amazon's pricing.Is it a nice gift? Sure. Is it practical enough to buy a couple of boxes for just ripping open packs and hoping for your favorite Eeveelution? At $59.99... jeesh, maybe not.Prismatic Evolution Stamped Surprise Box CardsUmbreon ex - 060/131 (Prismatic Evolutions Stamp)But still, it's been a popular pick amongst IGN's Daily Deals readers, so who am I to judge? Make your own decisions when buying, just don't say I didn't warn you. Otherwise, you can also consider picking up the singles you might be chasing via trusted retailers like TCGPlayer. Let's explore the top Prismatic Evolutions cards available to pick up today as well.Umbreon ex - 060/131$59.99 at TCG PlayerEevee ex - 075/131$17.25 at TCG PlayerFlareon ex - 014/131$14.95 at TCG PlayerJolteon ex - 030/131$13.95 at TCG PlayerVaporeon ex - 023/131$14.89 at TCG PlayerEspeon ex - 034/131$12.99 at TCG PlayerLeafeon ex - 006/131$12.99 at TCG PlayerGlaceon ex - 026/131$11.25 at TCG PlayerSylveon ex - 041/131 (Stamped)$12.82 at TCG PlayerIf you're, like me, just interested in the Suprise Box for the promo cards, it might be worth just buying the cards standlone. The most expensive (of course it's Umbreon) is going for the same price as one of these boxes, with the rest coming in at well under $20. So for the price of two to three Suprise Boxes, you can get the complete collection. If you want to rip open packs and pull these promos yourself then go for it, but it'll cost you.Prismatic Evolution Chase CardsUmbreon ex - 161/131Sylveon ex - 156/131Leafeon ex - 144/131Espeon ex - 155/131Vaporeon ex - 149/131Flareon ex - 146/131Glaceon ex - 150/131Jolteon ex - 153/131Roaring Moon ex - 162/131Eevee ex - 167/131Dropping over a grand on a Pokémon card sounds mental, but this Umbreon SIR has come down quite a bit after release. Sitting at $1100, that's a fall of around $500 and change. All the other chase cards of Prismatic Evolutions have fallen to under the $500 mark, with my favorite Roaring Moon ex SIR dropping to $221 at the time of writing. These prices may very well fall more as the year goes on and attention shifts, but these are pretty good prices considering what the bloated Pokémon card market would have you believe just a month ago.Pokémon TCG: Best Deals TodayPokémon TCG's Prismatic Evolutions expansion for Scarlet & Violet has been crazy popular, sending Pokémania 2025 over the edge due to tough pull rates and having few opportunities to rip the set open.But, if you're trying to keep up with all the other Pokémon TCG without getting ripped off or left behind, here's where to start, including the latest update on what can be bought here and now from trusted retailers in the US.Surging Sparks Booster BoxMimikyu ex Box$49.99 at AmazonShrouded Fable ETB$66.86 at AmazonPaldea Adventure Chest$57.00 at AmazonPokéball TIn 3 PackAzure Legends Tin$42.49 at AmazonHoliday 2024 Calendar$33.92 at Amazon2024 Trainer’s Toolkit$33.97 at AmazonDestined Rivals Booster BoxSee it at AmazonDestined Rivals ETBSee it at AmazonWant to know more about the best single Pokémon cards out there right now? My weekly Crashes and Climbers articles do just that. Want to pay less? The Glory of Team Rocket Japanese set singles are going for great prices right now too.Looking to preorder the new sets Destined Rivals or Black Bolt/White Flare? We've also got you covered, alongside every upcoming release for Pokémon TCG in 2025. Christian Wait is a contributing freelancer for IGN covering everything collectable and deals. Christian has over 7 years of experience in the Gaming and Tech industry with bylines at Mashable and Pocket-Tactics. Christian also makes hand-painted collectibles for Saber Miniatures. Christian is also the author of "Pokemon Ultimate Unofficial Gaming Guide by GamesWarrior". Find Christian on X @ChrisReggieWait.

TensorWave, a data center provider building facilities primarily with AMD hardware, has raised $100 million as it seeks to further build out its data center infrastructure. The funding round was led by Magnetar and AMD Ventures, and brings the company’s total capital raised to $146.7 million, according to Crunchbase. Maverick Silicon, Nexus Venture Partners and Prosperity7 also participated in the round. It’s a precarious time for data center projects. Tariff-related price hikes on components like server racks and chips could contribute to overall data center build costs increasing by 5% to 15%, per an analysis by TD Cowen. Investors are also wary of such projects accumulating too much capacity, particularly as the number of cheap AI services continues to grow. Overcapacity is reportedly one of the factors delaying OpenAI’s ambitious $500-billion Stargate data center project. Las Vegas, Nevada-based TensorWave claims that it hasn’t seen a slowdown in business, however. The company is on track to end the year with run-rate revenue of more than $100 million, which would mark a 20x increase from a year earlier, according to CEO Darrick Horton (pictured above; in the middle). Nvidia is the favored hardware vendor for data centers that are used for training and running AI models. But TensorWave embraced AMD early on, aiming to provide cloud services at lower prices. TensorWave recently deployed a “dedicated training” cluster of around 8,000 AMD Instinct MI325X GPUs. The new capital will enable the company to grow that cluster, as well as expand headcount and support “operational growth,” said Horton. Techcrunch event Join us at TechCrunch Sessions: AI Secure your spot for our leading AI industry event with speakers from OpenAI, Anthropic, and Cohere. For a limited time, tickets are just $292 for an entire day of expert talks, workshops, and potent networking. Exhibit at TechCrunch Sessions: AI Secure your spot at TC Sessions: AI and show 1,200+ decision-makers what you’ve built — without the big spend. Available through May 9 or while tables last. Berkeley, CA | June 5 REGISTER NOW TensorWave has a team of around 40 people at present, and expects headcount to reach over 100 by the end of the year. “This $100 million funding propels TensorWave’s mission to democratize access to cutting-edge AI compute,” Horton added. “Our 8,192 Instinct MI325X GPU cluster marks just the beginning as we establish ourselves as the emerging AMD-powered leader in the rapidly expanding AI infrastructure market.” Other data center providers placing bets on AMD’s AI chips range from startups like Lamini and Nscale to larger, more entrenched cloud providers such as Azure and Oracle. Horton co-founded TensorWave with Jeff Tatarchuk (pictured above; on the left) and Piotr Tomasik (pictured above; on the right) in 2023. Tatarchuk had previously launched cloud vendor VMAccel with Horton, and sold another startup, Lets Rolo, to digital identity firm LifeKey. Horton co-founded crypto mining company VaultMiner Technologies, VMAccel’s corporate parent. As for Tomasik, he co-launched influencer marketer site Influential, and is the second co-founder of Lets Rolo.

What just happened? It's another bad day to be a tech worker. Microsoft has announced that it is laying off 6,000 employees, around three percent of its entire workforce, globally. It brings the total number of tech industry layoffs in 2025 to almost 60,000 – and there's still more than half the year left. Microsoft said that is slashing three percent of its workforce across all levels, teams and geographies. The company employed around 228,000 workers as of the end of June, meaning about 6,000 people will be losing their jobs. Of that number, 1,985 layoffs will be tied to Microsoft's Redmond headquarters, including 1,510 in-office. While the cuts are affecting those at all levels, including at LinkedIn and international offices, the expected focus will be on reducing the number of management positions. Microsoft layoffs since 2023 Date (announcement) Reported head-count cut Main business areas hit 18 Jan 2023 10,000 Windows & devices, Xbox, HoloLens, recruiting & marketing 25 Jan 2024 1,900 Activision Blizzard, Xbox, ZeniMax after the ABK deal closed 3 – 4 Jun 2024 ≈1,000 (internal est.) Azure for Operators, HoloLens/mixed-reality, other "moon-shot" teams 12 Sep 2024 ≈650 Xbox publishing & game-studio support teams Jan – Feb 2025 ≈2,000 Company-wide "low-performer" cull (no severance in many cases) 13 May 2025 ≈6,000 (Microsoft says "< 3 % of staff") All geographies; focus on middle-management, LinkedIn In April, Microsoft Chief Financial Officer Amy Hood said on an earnings call that the company would be streamlining by "reducing layers with fewer managers." It was reported at the time that some Microsoft organizations were looking to increase their span of control, defined as the number of direct reports or subordinates a manager or supervisor oversees. The company also wanted to increase the number of coders compared to non-coders on projects. Regarding the latest round of layoffs, a Microsoft spokesperson said, "We continue to implement organizational changes necessary to best position the company for success in a dynamic marketplace." The move comes despite Microsoft beating expectations in its latest quarterly earnings report. The company made $70.1 billion in revenue for the quarter, up 13% from the same period last year, and it continues to invest billions of dollars in AI. The Windows maker has designated $80 billion in capital spending this fiscal year, with most of it aimed at expanding data centers for its artificial intelligence services. At the start of the year, Microsoft confirmed it was implementing performance-based layoffs, though it said those let go would be replaced with new hires. Soon after those performance cuts were revealed, the company said it was making more job cuts across its business, impacting employees in the gaming, experience & devices, sales, and security divisions. The company also tightened policies for low performers in April and added a voluntary exit plan. The latest round of cuts will be the largest since Microsoft laid off 10,000 people just over two years ago. Adding those to the 1,900 people let go after the Activision Blizzard deal closed, the Azure layoffs, the game studio cuts, and the others mentioned here brings the total to around 21,550 Microsoft layoffs since the start of 2023. // Related Stories

May 14, 2025Ravie LakshmananEndpoint Security / Vulnerability Microsoft on Tuesday shipped fixes to address a total of 78 security flaws across its software lineup, including a set of five zero-days that have come under active exploitation in the wild. Of the 78 flaws resolved by the tech giant, 11 are rated Critical, 66 are rated Important, and one is rated Low in severity. Twenty-eight of these vulnerabilities lead to remote code execution, 21 of them are privilege escalation bugs, and 16 others are classified as information disclosure flaws. The updates are in addition to eight more security defects patched by the company in its Chromium-based Edge browser since the release of last month's Patch Tuesday update. The five vulnerabilities that have come under active exploitation in the wild are listed below - CVE-2025-30397 (CVSS score: 7.5) - Scripting Engine Memory Corruption Vulnerability CVE-2025-30400 (CVSS score: 7.8) - Microsoft Desktop Window Manager (DWM) Core Library Elevation of Privilege Vulnerability CVE-2025-32701 (CVSS score: 7.8) - Windows Common Log File System (CLFS) Driver Elevation of Privilege Vulnerability CVE-2025-32706 (CVSS score: 7.8) - Windows Common Log File System Driver Elevation of Privilege Vulnerability CVE-2025-32709 (CVSS score: 7.8) - Windows Ancillary Function Driver for WinSock Elevation of Privilege Vulnerability While the first three flaws have been credited to Microsoft's own threat intelligence team, Benoit Sevens of Google Threat Intelligence Group and the CrowdStrike Advanced Research Team have been acknowledged for the discovery of CVE-2025-32706. An anonymous researcher has been credited with reporting CVE-2025-32709. "Another zero-day vulnerability has been identified in the Microsoft Scripting Engine, a key component used by Internet Explorer and Internet Explorer mode in Microsoft Edge," Alex Vovk, CEO and co-founder of Action1, said about CVE-2025-30397. "Attackers can exploit the flaw via a malicious web page or script that causes the scripting engine to misinterpret object types, resulting in memory corruption and arbitrary code execution in the context of the current user. If the user has administrative privileges, attackers could gain full system control – enabling data theft, malware installation, and lateral movement across networks." CVE-2025-30400 is the third privilege escalation flaw in DWM Core Library to be weaponized in the wild since 2023. In May 2024, Microsoft issued patches for CVE-2024-30051, which Kaspersky said was used in attacks distributing QakBot (aka Qwaking Mantis) malware. "Since 2022, Patch Tuesday has addressed 26 elevation of privilege vulnerabilities in DWM," Satnam Narang, senior staff research engineer at Tenable, said in a statement shared with The Hacker News. "In fact, the April 2025 release included fixes for five DWM Core Library elevation of privilege vulnerabilities. Prior to CVE-2025-30400, only two DWM elevation of privilege bugs were exploited as zero days – CVE-2024-30051 in 2024 and CVE-2023-36033 in 2023." CVE-2025-32701 and CVE-2025-32706 are the seventh and eighth privilege escalation flaws to be discovered in the CLFS component and have been exploited in real-world attacks since 2022. Last month, Microsoft revealed that CVE-2025-29824 was exploited in limited attacks to target companies in the U.S., Venezuela, Spain, and Saudi Arabia. CVE-2025-29824 is also said to have been exploited as a zero-day by threat actors linked to the Play ransomware family as part of an attack targeting an unnamed organization in the U.S., Broadcom-owned Symantec revealed earlier this month. CVE-2025-32709, likewise, is the third privilege escalation flaw in the Ancillary Function Driver for WinSock component to have come under abuse within a span of a year, after CVE-2024-38193 and CVE-2025-21418. It's worth noting that the exploitation of CVE-2024-38193 has been attributed to the North Korea-linked Lazarus Group. The development has prompted the U.S. Cybersecurity and Infrastructure Security Agency (CISA) to add all five vulnerabilities to its Known Exploited Vulnerabilities (KEV) catalog, requiring federal agencies to apply the fixes by June 3, 2025. Microsoft's Patch Tuesday update also addresses a privilege escalation bug in Microsoft Defender for Endpoint for Linux (CVE-2025-26684, CVSS score: 6.7) that could permit an authorized attacker to elevate privileges locally. Stratascale researcher Rich Mirch, who is one of the two researchers, acknowledged for reporting the vulnerability, said the issue is rooted in a Python helper script that includes a function ("grab_java_version()") to determine the Java Runtime Environment (JRE) version. "The function determines the location of the Java binary on disk by checking the /proc/<PID>/exe symbolic link and then executes the java -version command," Mirch explained. "The problem is the Java binary could be running from an untrusted location. A malicious local unprivileged user can create a process with the name java or javaw, which will eventually be executed with root privileges to determine the version of the JRE." Another notable flaw is a spoofing vulnerability affecting Microsoft Defender for Identity (CVE-2025-26685, CVSS score: 6.5) that allows an attacker with LAN access to perform spoofing over an adjacent network. "The lateral movement path detection feature can itself potentially be exploited by an adversary to obtain an NTLM hash," Adam Barnett, lead software engineer at Rapid7, said in a statement. "The compromised credentials in this case would be those of the Directory Services account, and exploitation relies on achieving fallback from Kerberos to NTLM." The vulnerability with the maximum-severity is CVE-2025-29813 (CVSS score: 10.0), a privilege escalation flaw in Azure DevOps Server that allows an unauthorized attacker to elevate privileges over a network. Microsoft said the shortcoming has been already deployed in the cloud and there is no action required on the part of customers. Software Patches from Other Vendors In addition to Microsoft, security updates have also been released by other vendors over the past few weeks to rectify several vulnerabilities, including — Found this article interesting? Follow us on Twitter  and LinkedIn to read more exclusive content we post. SHARE    